Deep learning models such as MLP, Transformer, and TCN have achieved remarkable success in univariate time series forecasting, typically relying on sliding window samples from historical data for training. However, while these models implicitly compress historical information into their parameters during training, they are unable to explicitly and dynamically access this global knowledge during inference, relying only on the local context within the lookback window. This results in an underutilization of rich patterns from the global history. To bridge this gap, we propose Predicting the Future by Retrieving the Past (PFRP), a novel approach that explicitly integrates global historical data to enhance forecasting accuracy. Specifically, we construct a Global Memory Bank (GMB) to effectively store and manage global historical patterns. A retrieval mechanism is then employed to extract similar patterns from the GMB, enabling the generation of global predictions. By adaptively combining these global predictions with the outputs of any local prediction model, PFRP produces more accurate and interpretable forecasts. Extensive experiments conducted on seven real-world datasets demonstrate that PFRP enhances the average performance of advanced univariate forecasting models by 8.4%.

Recent Large Reasoning Models (LRMs) excel at complex reasoning tasks but often suffer from overthinking, generating overly long and redundant reasoning trajectories.To explore its essence, our empirical analysis reveals that LRMs are primarily limited to recognizing task properties (i.e., difficulty levels) like humans before solving the problem, leading to a one-size-fits-all reasoning strategy.This observation motivates a fundamental question: Can we explicitly bootstrap such ability to alleviate overthinking in LRMs?To this end, we propose Think-How-to-Think (TH2T), a novel two-stage fine-tuning strategy that progressively inspires LRMs' difficulty cognition and redundancy cognition of LRMs.Specifically, we first inject Difficulty Dypnosis into output prefixes as cues for global, prospective reasoning strategy selection, stimulating the model's sharper sensitivity to task complexity and adaptive control of reasoning depth.Then, we incorporate Redundancy Hypnosis into inprogress reasoning steps, which serve as local, retrospective signals for behavior correction by identifying and eliminating superfluous reasoning detours.Experiments across 7B/14B/32B models demonstrate that TH2T significantly reduces inference costs by over 70% on easy tasks and 40% on complex ones without compromising performance.The resultant models exhibit a nascent ability for difficulty-aware reasoning, effectively mitigating behaviors like excessive reflection and looping, thereby paving the way for more cognitively efficient LRMs.

Large Language Models (LLMs) exhibit enhanced capabilities by Chain-of-Thought reasoning.However, the extended reasoning sequences introduce significant GPU memory overhead due to increased key-value (KV) cache.Existing KV cache compression methods mitigate memory bottlenecks but struggle in long reasoning tasks.In this paper, we analyze attention patterns in reasoning tasks and reveal a Token Importance Recurrence phenomenon: a large proportion of tokens regain high attention after multiple decoding steps, which is failed to capture by existing works and may lead to unpredictable eviction on such periodically critical tokens.To address this, we propose LazyEviction, an observation windowbased lagged eviction framework retaining latent recurring tokens by an eviction policy informed by token recurrence patterns.Extensive experiments demonstrate that LazyEviction reduces KV cache by 50%~70% while maintaining comparable accuracy, outperforming existing KV cache compression baselines.Our implementation code can be found at https: //github.com/Halo-949/LazyEviction.

Multimodal learning enables models to process and reason over diverse information sources, unlocking human-like perceptual and cognitive capabilities. As such models gain adoption, efficiently serving them on GPUs has become increasingly important. However, the modular architecture of multimodal models poses significant challenges to existing unimodal serving systems, which treat models as monolithic and overlook inter-module heterogeneity. This results in severe GPU underutilization. To address this, we propose Eevee, a multimodal serving system based on a new scheduling paradigm we call module multiplexing. Unlike prior approaches that execute all modules sequentially with uniform batch sizes, Eevee schedules modality-specific modules concurrently on the same GPU with independently tuned batching and resource allocation. This design enables fine-grained GPU sharing, boosting intra-GPU parallelism and improving request-level throughput. We implement a prototype of Eevee and evaluate it on several representative multimodal models (e.g., CLIP, BLIP, LLaVA, InternVL). Our results show that Eevee significantly outperforms state-of-the-art serving systems in both throughput and GPU utilization.

Large language models (LLMs) have driven significant progress in natural language processing, yet their training and fine-tuning remain limited by memory constraints, particularly the substantial memory footprints of optimizer states. Existing solutions address this challenge by offloading optimizer states and update tasks to the CPU, but this often leads to increased GPU idleness due to the CPU's limited computational capabilities, especially in pipeline parallelism.

Expert parallelism has become the prevailing paradigm to serve Mixture-of-Experts (MoE) models. Its efficiency depends on the communication and computation latencies of the GPUs, which are linked to the placement of experts in the GPUs. Existing works for optimizing expert placement focus on leveraging past requests' expert activation patterns. However, they demonstrate deficiencies facing diverse and rapidly changing request patterns, calling for an online, proactive approach. Implementing such an approach requires addressing several challenges: the uncertainty associated with incoming requests' expert activation, the cost of expert migration, and the NP-hard complexity in optimization. Therefore, we present Director, a new distributed MoE serving system that minimizes end-to-end latency via prediction-driven, online expert placement. Director uses either a lightweight cascaded predictor or a low-bit quantized replica for expert activation patterns of incoming requests. An online migration module then enacts the changes with near-zero downtime by executing migrations in compute-bound phases, keeping disruption bounded. At its core, a relaxation-based expert placement optimizer operates under capacity constraints, runs in polynomial time, and achieves a (1+ ϵ) approximation ratio. Finally, we implement a prototype and demonstrate, through extensive experiments, a reduction in end-to-end latency of 11 ~ 55% for popular MoE models (e.g., Mistral, DeepSeek and Qwen) compared to existing work.

Zero-shot skeleton action recognition endeavors to classify novel action categories by transferring previously learned seen skeleton-semantic priors to unseen categories. However, current methods struggle to distinguish highly similar action categories, primarily due to the coarse-grained cross-modal alignment and non-discriminative representation space. To address these issues, we proposeSTAR++, a novel framework that aligns skeleton and semantics in a fine-grained and conditional manner. The key idea is to first establish region-level correspondences between body parts and semantic cues, and then utilize these local alignments to inform a global alignment process. This design is inspired by human visual cognition, which first attends to crucial local details before perceiving the broader scene. Concretely, we refine both skeleton and semantic representations with a dual-prompt attention mechanism driven by the structural decomposition of the human body and side information generated by a large language model (LLM). This encourages skeleton representations to be more compact within each class and semantic embeddings to be more separable across classes, which helps resolve ambiguity between highly similar actions and provides better interpretability of how unseen actions are perceived. Furthermore, we construct a region-aware holistic fusion module that aggregates these fine-grained features into a unified representation, yielding more discriminative holistic representations. Finally, the global alignment is conditioned on region-aware semantics feedback derived from fine-grained alignment, forming a conditional process that achieves more effective cross-modal alignment. Extensive experiments on four mainstream benchmarks demonstrate that our method achieves state-of-the-art performance in the zero-shot learning (ZSL) and generalized zero-shot learning (GZSL) settings.

Text-guided video inpainting faces a unique challenge: while textual prompts enable semantic control, diffusion model often generate content that is spatially incompatible with the surrounding visual context, producing correct but incongruous results. This spatial disharmony, combined with the inherent temporal consistency requirements of video inpainting, creates a complex optimization landscape. We present Flow-based Efficient Diffusion (FloED), a framework that harmonizes text-guided generation with spatial-temporal coherence through optical flow guidance. Our key insight is that optical flow encodes the underlying scene dynamics, serving as a contextual bridge between semantic generation and environmental constraints. FloED employs a dual-branch architecture: a time-agnostic flow completion branch that captures scene-consistent motion patterns, and a main inpainting branch where multi-scale flow adapters inject this contextual information at different semantic levels. Besides, we introduce a flow-guided denoising acceleration that leverages motion information to reduce redundant computations during the multi-step diffusion process. Combined with our flow attention caching mechanism, FloED efficiently minimizes the computational overhead typically associated with optical flow integration. Extensive experiments demonstrate that FloED significantly improves spatial compatibility (4.9 dB PSNR gain) while maintaining superior efficiency, temporal stability and text alignment, establishing new state-of-the-art performance on both object removal and background restoration tasks.

Buy Now Pay Later (BNPL) is a rapidly proliferating e-commerce model, offering consumers to get the product immediately and defer payments. Meanwhile, emerging blockchain technologies endow BNPL platforms with digital currency transactions, allowing BNPL platforms to integrate with digital wallets. However, the transparency of transactions causes critical privacy concerns because malicious participants may derive consumers' financial statuses from on-chain asynchronous payments. Furthermore, the newly created transactions for deferred payments introduce additional time overhead, which weakens the scalability of BNPL services. To address these issues, we propose an efficient and privacy-preserving blockchain-based asynchronous payment scheme (Epass), which has promising scalability while protecting the privacy of on-chain consumer transactions. Specifically, Epass leverages locally verifiable signatures to guarantee the privacy of consumer transactions against malicious acts. Then, a privacy-preserving asynchronous payment protocol is further constructed by leveraging time-release encryption to control trapdoors of the redactable blockchain, reducing time overhead by modifying transactions for deferred payment. We give formal definitions and security models, generic structures, and formal proofs for Epass. Extensive comparisons and experimental analysis show that Epass achieves KB-level communication costs, and reduces time overhead by more than four times in comparisons with locally verifiable signatures and Go-Ethereum private test networks.

Sharding is considered a promising solution for scaling blockchain systems. However, most existing sharding systems have not considered the dynamics of the environment when making a sharding strategy, including the change of pending transactions, the leaving and joining of participants, and malicious attacks, which could cause performance instability and security issues. To address it, in this paper, we propose an intelligent and efficient dynamic sharding technology to advance the blockchain system performance and security. We first propose a formal and general evaluation framework for blockchain sharding in a dynamic environment, and conclude an optimization target for the system performance and security. To achieve a long-term benefit for the optimization target, a deep reinforcement learning (DRL)-based sharding approach has been proposed to intelligently make optimal sharding strategies. Next, we propose an adaptive resharding protocol to efficiently reduce the overhead introduced by dynamic sharding. Our experimental results illustrate that our proposed dynamic sharding in a simulation testbed can achieve 2.8 times transactions per second compared to traditional static sharding systems, and guarantee high security in a dynamic environment.

The growing demand for computing power drives more artificial intelligence (AI) model owners to outsource their models to public clouds, relying on cloud servers to manage which users can use, train, or upgrade AI models. Unfortunately, existing work cannot simultaneously manage the entire model lifecycle in public clouds when considering model stealing attacks, where cloud servers covertly replicate AI models and deliver AI services to unauthorized users for profit. As a result, model owners are forced to conduct training and upgrades in private environments before deploying models to clouds for service delivery, which poses significant challenges in collaboration and maintenance, particularly for models requiring frequent upgrades. In this paper, we introduce$\mathsf {Trident}$, the first secure cloud-based framework for flexible AI model lifecycle management, including availability, trainability, and upgradability. By leveraging multiple cryptographic techniques, such as access control trees,$\mathsf {Trident}$ensures that AI models and their management policies are tightly coupled, compelling cloud servers to execute only specified model operations without violating management policies, thereby resisting model stealing attacks. Rather than straightforward cryptographic applications, we address a series of technical challenges, including shifting the focus of access control trees from data to model management and maintaining downward-compatible model management rights. We propose two detailed constructions: Semi-$\mathsf {Trident}$and Full-$\mathsf {Trident}$, tailored for semi-delegation and full-delegation scenarios, i.e., whether model owners need to interact with cloud servers while delivering AI services. Theoretical complexity analysis and security analysis prove the competitive efficiency and security. Experimental results show that compared to assembling existing partial-function schemes, Semi-$\mathsf {Trident}$and Full-$\mathsf {Trident}$achieve around$3.6\times$improvement in time costs and$5\times$improvement in communication overhead.

The transition to Industry 5.0, supported by the high-bandwidth and low-latency capabilities of 6G networks, accelerates the demand for personalized products, making efficient and accurate product style recognition a critical task. However, traditional recognition models, which rely heavily on large-labeled datasets, struggle with generalization and efficiency in such data-rich environments. In contrast, multimodal large language models (MLLMs), such as Qwen-VL, offer new possibilities by integrating visual and textual inputs, requiring less annotated data and demonstrating strong generalization. Nevertheless, most existing MLLMs are general-purpose and lack optimized prompts tailored for domain-specific tasks, limiting their recognition performance. To address this, we propose a prompt optimization framework that combines genetic algorithms (GA) and chain of thought (CoT) reasoning. GA is used to iteratively refine the instruction component of prompts, while CoT encourages step-by-step reasoning, improving the model’s interpretability and recognition accuracy. Validated on a custom-built apparel dataset and the public FashionStyle14 dataset, our method enables Qwen-VL to achieve significant accuracy improvements of 8.7% and 19.6%, respectively, over the baseline. This study makes three contributions: a framework to adapt general MLLMs for fine-grained style recognition; a demonstration of 6G-enabled real-time artificial intelligence reasoning loops; and a high-quality, multidimensional style dataset for future research. This demonstrates the effectiveness and stability of the proposed framework in handling fine-grained style recognition tasks, contributing to the development of intelligent design systems for Industry 5.0.

Accurate recognition of early-stage pneumoconiosis presents significant challenges due to the irregular morphology, diffuse distribution, and small size of pulmonary lesions. Existing 2D methods struggle to focus on lesion-level 3D characteristics and inter-slice correlations in localized weak lesion regions, resulting in incomplete feature extraction and inaccurate calculation of lesion volume. To obtain complete 3D fine-grained lesion features in the entire lung, this paper proposes an early pneumoconiosis recognition network (EPRNet) to enhance fine-grained feature acquisition abilities and discover inter-slice correlations, thereby improving early pneumoconiosis recognition accuracy in a more structured and flexible manner. Specifically, to obtain the fine-grained 3D features of early pneumoconiosis more comprehensively, a distance-similarity graph encoding module is proposed to construct and encode the relationships of the distributed tiny lesions within CT slices, integrating the spatial positions and the corresponding feature similarities of the lesions to improve the accuracy of pneumoconiosis feature representation. To adaptively preserve accurate graph representations of the correlations of the lesions across inter-CT slices, a hierarchical dynamic-scored adaptive pooling module is proposed to discover the potential long distance correlations between cross-slices, obtaining spatial semantic information of diffused lesions in the entire lung. Experimental results based on multiple datasets demonstrate that EPRNet achieves state-of-the-art performance while exhibiting better generalization. The ablation experiments also prove the effectiveness of each module in EPRNet.

Ultrasound imaging and biochemical examinations are the primary methods for diagnosing Hashimoto's thyroiditis (HT). However, neither of them is sufficient to accurately diagnose HT alone. Most existing multimodal models for HT diagnosis focus primarily on extracting and concatenating features from different modalities, which are ineffective due to the dimensional imbalance of the features between the textual and image data. To address this issue, we propose a novel Multimodal Collaborative Fusion Learning (MCFL) approach, which can enhance and recalibrate the biochemical indicators using ultrasound images, effectively improving the significance and specificity of biochemical indicators for the diagnosis of HT. Specifically, MCFL first constructs a novel INNet to convert the image-level characteristics of the HT ultrasound image into two numerical indicators, i.e., the Local prominent inflammatory (Lpi) and the Global diffuse lesion (Gdl), unifying image data and textual data into a single representation space. Then, a decision tree-based optimization strategy is employed to supervise the training of INNet, interactively recalibrating biochemical indicators with the guidance of the two numerical indicators mentioned above and obtaining a more accurate feature representation of HT. Finally, based on the deep Q-learning framework, a reward mechanism is established to guide the HT diagnostic process, in which the experience replay mechanism and the $\epsilon $ -greedy strategy are utilized collaboratively to improve the accuracy and robustness of the model. Extensive experiments are conducted on a multimodal dataset from multiple medical centers, and the results demonstrate that MCFL achieves state-of-the-art performance, setting a new benchmark.

Due to its capability to ensure communication security and improve spectral efficiency, Simultaneously Transmitting and Reflecting Reconfigurable Intelligent Surface (STAR-RIS)-assisted covert communications in Rate-Splitting Multiple Access (RSMA) networks have drawn widespread attention. However, existing studies often overlook the impact of the mobility of wardens and users on long-term covert performance, and lack joint optimization of covert rate and energy consumption. Meanwhile, introducing STAR-RIS significantly increases system parameter dimensionality, rendering traditional methods inefficient. Although reinforcement learning offers advantages, it still faces challenges in training efficiency and resource overhead within high-dimensional state-action spaces. Therefore, this paper proposes a quantum reinforcement learning-based algorithm, named QUERC, for STAR-RIS-assisted covert communications in RSMA networks. Specifically, we first formulate a long-term covert energy efficiency maximization problem under a dynamic environment with mobile users and wardens. Then, we propose a novel hybrid quantum neural network architecture to solve this problem. This architecture integrates a fully connected layer, a variational quantum circuit, an action post-processing mechanism, and a regularized objective, enhancing policy stability and generalization in high-dimensional action spaces. Finally, extensive experiments demonstrate that, compared with TPG, QPG, RIS-NOMA, RPS, and greedy approaches, QUERC algorithm achieves superior average covert energy efficiency and offers significant advantages in computational and inference performance.

Low Earth Orbit (LEO) satellite networks are expected to become a key component of Sixth Generation (6G) communication networks, to relieve the communication burden on ground networks. Driven by the rapid advancement of communication technologies and intelligent applications, dense traffic flow in the Internet of Vehicles (IoV) inevitably leads to a surge in task generation and an increased demand for network resources. The requirement for low latency further intensifies this challenge, making it difficult to rely solely on ground network resources to process tasks efficiently and promptly. Conversely, relying only on satellite networks for task processing results in high costs. Therefore, flexibly integrating LEO satellite links based on real-time traffic conditions, task demands, and the real-time state of ground network resources becomes an effective solution. However, achieving such a goal poses significant challenges in efficient allocation and balance between ground and LEO satellite network resources. Therefore, we propose a dynamic multi-task partial offloading algorithm based on LEO satellite-ground network collaboration to efficiently allocate resources between ground and satellite networks in real time. We first introduce the utility gain as a metric to evaluate task scheduling preference and design an improved iterative algorithm to jointly optimize the offloading ratio and channel allocation to maximize system utility. Finally, based on the real-world dataset of Shanghai (China), we demonstrate the significant advantages of the proposed strategy over representative methods in terms of delay, vehicle satisfaction, and system utility.

Although AI has been extensively adopted and has profoundly transformed our lives, it is not feasible to directly deploy large AI models on edge devices with limited resources. To enhance the performance of Edge Intelligence (EI), model split inference has been proposed. In this approach, an AI model is segmented into sub-models, with the most resource-intensive parts offloaded wirelessly to the edge server. This reduces the resource demands and inference latency on the device. However, previous studies have primarily focused on enhancing and optimizing system Quality of Service (QoS), often overlooking Quality of Experience (QoE), which is another crucial aspect for users. Even though QoE has been extensively studied in Edge Computing (EC), the distinct differences between task offloading in EC and split inference in EI, along with specific QoE issues that remain unaddressed in both fields, render these algorithms ineffective for edge split inference scenarios. Therefore, this paper introduces an effective resource allocation algorithm, dubbed ERA, which aims to: 1) expedite split inference in EI, and 2) balance inference delay, QoE, and resource consumption. ERA incorporates resource consumption, QoE, and inference latency to determine the most optimal model split and resource allocation strategies. Given that it is impossible to simultaneously minimize inference delay and resource consumption while maximizing QoE, we employ a gradient descent-based algorithm to find the best possible compromise. Furthermore, to address the complexity arising from parameter discretization in the gradient descent algorithm, we have developed a pipeline gradient descent approach, known as PipGD. We have also examined the properties of the proposed algorithms, including their convergence, complexity, and approximation error. The experimental results clearly show that ERA outperforms previous studies significantly in terms of performance.

The rise of pre-trained foundation models (FMs) has popularized the trend of fine-tuning FMs to fit downstream tasks, while Federated Learning (FL) has become the de-facto approach for training distributed data with privacy-preservation. However, fine-tuning FMs in FL faces overwhelming overheads due to its bulky nature. While freezing parameters in FM have the potential to accelerate FL training, existing freezing strategies statically freeze parameters on specified or already converged layers, incur severe accuracy degradation, and resource-inefficiency in heterogeneous environments. In this paper, we propose AFedLF, an adaptive freezing framework for FM in FL, to accelerate its wall-clock time for convergence without losing its final accuracy. However, this poses great challenges, as different freezing strategies lead to different accuracy gains and time overheads, while unfreezing more layers may bring marginal accuracy gains but significant time overheads. To address this challenge, AFedLF mathematically establishes a correlation between the freezing strategy and the accuracy gain and time overhead, and allocates adaptive freezing strategies to clients, based on our insight that unfreezing more layers on devices with strong computation and communication capabilities helps improve resource efficiency. Besides, AFedLF incorporates our well-designed intermediate result caching scheme with constant approximation ratios utilizing the limited storage capacity on mobile devices to cache intermediate results to skip forward propagation, further saving wall-clock time. Finally, we implemented AFedLF using an open-source FL benchmark, and extensive trace-driven experimental results showed that AFedLF accelerates wall-clock time by up to 6.1× compared to state-of-the-art solutions, without sacrificing accuracy.

Nowadays, mobile cloud storage has become increasingly prevalent for processing mobile data due to its convenience and resources. Towards the right to restriction of processing in current data regulations, some solutions have been proposed to empower data owners with identity-based hierarchical control for equality, comparison, and plaintext analytics operations. However, existing solutions either rely on specific hardware environments (i.e., trusted execution environments) or face two significant issues: identity privacy breaches, where attackers can identify targeted users, and excessive overhead in multi-user scenarios, as each user requires a distinct ciphertext. In this paper, we leverage multiple cryptographic primitives to introduce PODS, an efficient and secure hierarchical data processing control scheme for multi-user mobile cloud storage. PODS allows the data owner to generate a single ciphertext for multiple users without hardware reliance. Technically, we reconstruct identity-based encryption by leveraging well-designed common and private parameters, thereby shifting the focus from data itself to data processing operations. Then, we encode multiple targeted identities as polynomial coefficients and integrate these coefficients into identity-based data processing control. We achieve threefold benefits, effectively reducing overhead, hierarchically processing users' rights, and concealing the identities of targeted users to protect privacy. Security analysis proves the security of PODS. Experiments demonstrate that PODS achieves around$8\times$and$37\times$improvement in computation and communication compared to existing related works.

Ultrasound imaging can reveal the typical changes in thyroid tissue caused by Hashimoto's Thyroiditis (HT), which plays a crucial role in HT diagnosis. Clinicians should perform ultrasound imaging from multiple anatomical planes to obtain a series of thyroid images, thereby collecting comprehensive information about HT lesions from different perspectives. However, ultrasound images lacking typical HT characteristics or containing extensive non-HT regions may hinder the identification of localized lesions, thereby increasing the complexity of diagnosis. In addition, the ability to capture synergistic interactions among multiple lesions in critical ultrasound images is crucial for improving the diagnostic accuracy of HT. To address these challenges, a novel weakly supervised multi-instance learning model, HTMIL, is proposed for HT diagnosis, requiring only patient-level data annotation. HTMIL consists of a Global-Neighbor extraction Layer (GNL) and a Cross-Aggregation Layer (CAL). Specifically, a Global-Neighbor Awareness (GNA) module is proposed in GNL to allow HTMIL to focus on key localized lesions and ignore noise from unrelated regions in HT ultrasound images, thus improving the effectiveness of extracting focused features from localized lesions. The Adaptive Bilinear Pooling (ABP) module is introduced in CAL to capture the interaction features between key localized lesions and other characteristics of HT in critical ultrasound images, thus achieving synergistic diagnosis and further increasing diagnostic accuracy. HTMIL achieves state-of-the-art (SOTA) performance with 87.67% accuracy and 91.73% AUC on a multicenter HT dataset, and its robustness is further validated on a public dataset.

Few-shot knowledge graph completion (FKGC) aims to infer missing triples for long-tail relationships using a small set of References. Existing FKGC models focus mainly on entity representation aggregation, heavily relying on interactions between central entities and their neighbors. However, real-world knowledge graphs contain relations with multiple semantics, and existing models struggle to capture the diverse semantic information of the relations in different contexts. To address this issue, we propose a novel FKGC model, context-aware relational learning and multidimensional matching (CRL-MM). First, CRL-MM enhances the representation of task relations by obtaining semantic information in different scenarios based on the semantic similarity between task relations and background relations. Second, unlike previous models, which rely mainly on neighborhood relations to capture relation information, CRL-MM considers the entity pair and its neighborhood as a unified contextual whole, aggregating neighborhood information through adaptive task relations and paired entity awareness to improve entity encoding. In addition, during the matching phase, we design a matching network from multiple dimensions, which includes not only the similarity score of the entity pairs but also the triple rationality score to further improve the generalizability of the model. Extensive experiments on public benchmark datasets show that CRL-MM outperforms state-of-the-art methods, and the ablation experiments also demonstrate the effectiveness of each module of the proposed CRL-MM.

The evolution of 6G networks introduces heterogeneous services with stringent computing and latency demands. However, constrained edge resources, intricate task dependencies, and dynamic network fluctuations intensify resource contention, increasing the risk of node faults and service interruption. Current fault-tolerant methodologies lack the necessary adaptability to handle the coupled complexity of task interdependencies and volatile resource states, leading to sub-optimal decisions or excessive system overhead. To address these challenges, this paper innovatively proposes TransDiffuse—an intelligent preemptive service migration framework for 6G edge networks. First, the framework employs a Transformer-GAT hybrid model to capture long-range temporal load dynamics and spatial topological constraints, enabling accurate failure prediction. Second, to navigate the trade-off between migration overhead and service robustness, we devise a diffusion-based decision module. This module efficiently explores the discrete combinatorial solution space to synthesize near-optimal service orchestration. Furthermore, a comprehensive evaluation system is constructed to validate the effectiveness of TransDiffuse. Experiments demonstrate that TransDiffuse reduces energy consumption by 32.4%, decreases task completion time by 25.6%, and improves resource balance by 18.7%, while keeping service violations below 5%. This work achieves joint optimization of energy, delay, and resource efficiency, offering a robust solution for resilient service orchestration in 6G edge networks.

Blockchain sharding has been deemed a promising solution that can substantially improve blockchain scalability. However, developers must overcome two major technical challenges to implement a sharded blockchain. The first challenge is the high cross-shard transaction ratio in blockchain shards. This issue significantly degrades the throughput of a sharded blockchain. The second challenge is the imbalanced workloads across blockchain shards. In a blockchain with imbalanced workloads, some busy shards have to handle an overwhelming number of transactions and thus become congested. Facing these two challenges, a dilemma is that it is difficult to guarantee a lowcross-shard transaction ratioand maintain thebalanced workloadsacross all shards, simultaneously. We believe that a fine-grained account allocation strategy can address this dilemma. To this end, we formulate the tradeoff between these two metrics as a network-partition problem. We then solve this problem by proposing a sharding protocol, namedShardCutter, which includes the following two crucial components: a community-aware account partition algorithm and a fine-tuned account migration mechanism. Finally, experimental results demonstrate that the proposed protocol outperforms other baselines in terms of throughput, makespan, cross-shard transaction ratio, and the workload balance of shards’ transaction pool.

Recently, federated learning (FL) has gained momentum because of its capability in preserving data privacy. To conduct model training by FL, multiple clients exchange model updates with a parameter server over the Internet. To accelerate the communication speed, it has been explored to deploy a programmable switch (PS) in lieu of the parameter server to coordinate clients. The challenge to deploy the PS in FL lies in its scarce memory space, prohibiting running memory consuming aggregation algorithms on the PS, like TopK. To overcome this challenge, we propose Federated Learning In-Network Aggregation with Global TopK Model Updates (Fed-GTopK) algorithm, consisting of two phases: voting and aggregating. In the voting phase, clients efficiently upload their votes for top model updates to the PS for selecting global top ones. Note that the voting phase consumes little memory or communication resources by only exploring the sparsity of top model updates without transmitting any model update values. In the aggregating phase, clients can reach the consensus to upload global top model updates such that the PS can swiftly aggregate global top model updates in a streamline manner without consuming much memory cost. Compared with existing works, our study is the first one accelerating in-network aggregation for FL by sparsifying model updates, and hence achieving the highest compression rate and the best learning performance. Finally, we conduct extensive experiments by using public datasets to demonstrate that Fed-GTopK remarkably surpasses the state-of-the-art baselines in terms of both model accuracy and communication traffic.

Abstract Hypergraph decomposition is a fundamental problem in hypergraph analysis which breaks down hypergraphs into cohesive subgraphs and functional units with dense interactions. Hyperedge intersections and overlaps capture the unique property of shared elements (vertices) between groups (hyperedges) in hypergraphs, revealing cohesive substructures not apparent when focusing solely on individual connections. Despite the significance of hyperedge overlap as a measure of hypergraph cohesiveness, existing models for hypergraph decomposition fail to capture this feature. In this paper, we study the problem of hypergraph decomposition with intersection bounds. We propose the ( k , s )-core, a new cohesive subgraph model incorporating both a vertex degree constraint k and a hyperedge intersection constraint s . This model includes two types: (1) strong ( k , s )-cores, where connected hyperedges share at least s vertices, enforcing strong hyperedge overlap, and (2) weak ( k , s )-cores, where hyperedges are connected through s -walks, allowing for a looser overlap. We prove that our definition of ( k , s )-cores exhibits uniqueness and hierarchical properties. Based on the properties, we develop two decomposition algorithms: a bottom-up algorithm for strong ( k , s )-cores, which uses a heuristic hyperedge removal mechanism to maintain consistent decomposition results and employs a union-find data structure for efficient connectivity identification, and a top-down algorithm for weak ( k , s )-cores that preserves the subgraph containment relationship. Our algorithms achieve traversal efficiency by processing each hyperedge in the hypergraph only once. Additionally, all ( k , s )-cores can be efficiently stored with minimal memory overhead. Comprehensive experiments and case studies show that the ( k , s )-core model outperforms existing methods in capturing cohesive subgraphs with overlaps in hypergraphs. Furthermore, the proposed algorithms demonstrate high efficiency and scalability, making them well-suited for real-world hypergraphs.

Perturbation-based mechanisms, such as differential privacy, mitigate gradient leakage attacks by introducing noise into the gradients, thereby preventing attackers from reconstructing clients' private data from the leaked gradients. However, can gradient perturbation protection mechanisms truly defend against all gradient leakage attacks? In this paper, we present the first attempt to break the shield of gradient perturbation protection in Federated Learning for the extraction of private information. We focus on common noise distributions, specifically Gaussian and Laplace, and apply our approach to DNN and CNN models. We introduce Mjölnir, a perturbation-resilient gradient leakage attack that is capable of removing perturbations from gradients without requiring additional access to the original model structure or external data. Specifically, we leverage the inherent diffusion properties of gradient perturbation protection to develop a novel diffusion-based gradient denoising model for Mjölnir. By constructing a surrogate client model that captures the structure of perturbed gradients, we obtain crucial gradient data for training the diffusion model. We further utilize the insight that monitoring disturbance levels during the reverse diffusion process can enhance gradient denoising capabilities, allowing Mjölnir to generate gradients that closely approximate the original, unperturbed versions through adaptive sampling steps. Extensive experiments demonstrate that Mjölnir effectively recovers the protected gradients and exposes the Federated Learning process to the threat of gradient leakage, achieving superior performance in gradient denoising and private data recovery.

This paper investigates a challenging problem of zero-shot learning in the multi-label scenario (MLZSL), wherein the model is trained to recognize multiple unseen classes within a sample (e.g., an image) based on seen classes and auxiliary knowledge, e.g., semantic information. Existing methods usually resort to analyzing the relationship of various seen classes residing in a sample from the dimension of spatial or semantic characteristics and transferring the learned model to unseen ones. However, they neglect the integrity of local and global features. Although the use of the attention structure will accurately locate local features, especially objects, it will significantly lose its integrity, and the relationship between classes will also be affected. Rough processing of global features will also directly affect comprehensiveness. This neglect will make the model lose its grasp of the main components of the image. Relying only on the local existence of seen classes during the inference stage introduces unavoidable bias. In this paper, we propose a novel and comprehensive visual-semantic framework for MLZSL, dubbed Epsilon, to fully make use of such properties and enable a more accurate and robust visual-semantic projection. In terms of spatial information, we achieve effective refinement by group aggregating image features into several semantic prompts. It can aggregate semantic information rather than class information, preserving the correlation between semantics. In terms of global semantics, we use global forward propagation to collect as much information as possible to ensure that semantics are not omitted. Experiments on large-scale MLZSL benchmark datasets NUS-Wide and Open-Images-v4 demonstrate that the proposed Epsilon outperforms other state-of-the-art methods with large margins.

Recent years have witnessed the rise of Neural-enhanced Video Streaming (NeVS), which integrates neural restoration models into video codecs for higher compression-restoration performance. Despite its benefit, existing work has not well explored the full potential of NeVS paradigm, due to: (1) post-streaming restoration by decoder while lacking the proactive collaboration of encoder, (2) end-to-end optimization based on conventional rate-distortion theory, which has been verified that low distortion is not always a synonym for high perceptual quality, and (3) coupled design for domain-specific tasks that cannot generalize to various video codecs. Observing these limitations, our objective is not to incrementally present an improved restoration model. Instead, we focus on the encoder-decoder synergy, i.e., the codec, which is non-trivial since it inherently strikes the rate-distortion-perception trade-off of NeVS. Aiming at this target, we propose the Diffusion-enhanced Neural Codec (DeNC), a plug-and-play module for current NeVS paradigm, to significantly reduce the required bitrates while preserving high perceptual quality of restored videos. Our key design is twofold. First, DeNC improves the encoder's compression efficiency by simultaneously reducing the resolution and color bit-depth of frame referencing. Second, DeNC empowers the decoder with perception-oriented restoration capability by making its diffusion-based restoration process aware of the encoder's compression conditions. Real-world evaluations show that DeNC improves compression ratios with nearly an order of magnitude and achieves much higher restoration quality (e.g., 93+ VMAF and 23% higher MOS) over the latest baselines.

Zero-shot skeleton action recognition is a non-trivial task that requires robust unseen generalization with prior knowledge from only seen classes and shared semantics. Existing methods typically build the skeleton-semantics interactions by uncontrollable mappings and conspicuous representations, thereby can hardly capture the intricate and fine-grained relationship for effective cross-modal transferability. To address these issues, we propose a novel dyNamically Evolving dUal skeleton-semantic syneRgistic framework with the guidance of cOntext-aware side informatioN (dubbed Neuron), to explore more fine-grained cross-modal correspondence from micro to macro perspectives at both spatial and temporal levels, respectively. Concretely, 1) we first construct the spatial-temporal evolving micro-prototypes and integrate dynamic context-aware side information to capture the intricate and synergistic skeleton-semantic correlations step-by-step, progressively refining cross-model alignment; and 2) we introduce the spatial compression and temporal memory mechanisms to guide the growth of spatial-temporal micro-prototypes, enabling them to absorb structure-related spatial representations and regularity-dependent temporal patterns. Notably, such processes are analogous to the learning and growth of neurons, equipping the framework with the capacity to generalize to novel unseen action categories. Extensive experiments on various benchmark datasets demonstrated the superiority of the proposed method1.

Despite the significant advancements, existing object removal methods struggle with incomplete removal, incorrect content synthesis and blurry synthesized regions, resulting in low success rates. Such issues are mainly caused by the lack of high-quality paired training data, as well as the self-supervised training paradigm adopted in these methods, which forces the model to in-paint the masked regions, leading to ambiguity between synthesizing the masked objects and restoring the background. To address these issues, we propose a semi-supervised learning strategy with human-in-the-loop to create high-quality paired training data, aiming to train a Robust Object Remover (RORem). We first collect 60K training pairs from open-source datasets to train an initial object removal model for generating removal samples, and then utilize human feedback to select a set of high-quality object removal pairs, with which we train a discriminator to automate the following training data generation process. By iterating this process for several rounds, we finally obtain a substantial object removal dataset with over 200K pairs. Fine-tuning the pre-trained stable diffusion model with this dataset, we obtain our RORem, which demonstrates state-of-the-art object removal performance in terms of both reliability and image quality. Particularly, RORem improves the object removal success rate over previous methods by more than 18%. The dataset, source code and trained model are available at https://github.com/leeruibin/RORem.

With the extraordinary success of generative artificial intelligence, large pretrained models (LPMs) have been widely used to achieve human-level performance. Despite the one-shot capability, it is always preferred to fine-tune the LPMs for domain-specific downstream tasks. Therefore, the federated learning system is leveraged to fine-tune the large pretrained models enabling concurrrently use multiple distributed clients as well as their local datasets. While the first-order fine-tuning methods suffer from high computational and memory costs due to the backward propagation, we are motivated to propose a federated zeroth-order fine-tuning method with only forward propagation. Moreover, we also leverage differential privacy to further preserve the data privacy of local clients. Experimental results illustrate that our proposed federated zeroth-order method can reduce the memory and retain a similar testing accuracy over the state-of-the-art benchmarks.

This paper presents Variables-Adaptive Mixture of Experts (VA-MoE), a novel framework for incremental weather forecasting that dynamically adapts to evolving spatiotemporal patterns in real-time data. Traditional weather prediction models often struggle with exorbitant computational expenditure and the need to continuously update forecasts as new observations arrive. VA-MoE addresses these challenges by leveraging a hybrid architecture of experts, where each expert specializes in capturing distinct sub-patterns of atmospheric variables (e.g., temperature, humidity, wind speed). Moreover, the proposed method employs a variable-adaptive gating mechanism to dynamically select and combine relevant experts based on the input context, enabling efficient knowledge distillation and parameter sharing. This design significantly reduces computational overhead while maintaining high forecast accuracy. Experiments on real-world ERA5 dataset demonstrate that VA-MoE performs comparable against state-of-the-art models in both short-term (e.g., 1–3 days) and long-term (e.g., 5 days) forecasting tasks, with only about 25\% of trainable parameters and 50\% of the initial training data.

Arbitrary resolution image generation provides a consistent visual experience across devices, having extensive applications for producers and consumers. Current diffusion models increase computational demand quadratically with resolution, causing 4K image generation delays over 100 seconds. To solve this, we explore the second generation upon the latent diffusion models, where the fixed latent generated by diffusion models is regarded as the content representation and we propose to decode arbitrary resolution images with a compact generated latent using a one-step generator. Thus, we present the \textbf{InfGen}, replacing the VAE decoder with the new generator, for generating images at any resolution from a fixed-size latent without retraining the diffusion models, which simplifies the process, reducing computational complexity and can be applied to any model using the same latent space. Experiments show InfGen is capable of improving many models into the arbitrary high-resolution era while cutting 4K image generation time to under 10 seconds.

In the field of AI security, the vulnerability of deep neural networks has garnered widespread attention. Specifically, the sensitivity of DNNs to adversarial examples (AEs) can lead to severe consequences, even small perturbations in input data can result in incorrect predictions. AEs demonstrate transferability across models, however, targeted attack success rates (TASRs) remain low due to significant differences in feature dimensions and decision boundaries. To enhance the transferability of targeted AEs, we propose a novel approach by introducing Inverse Target Gradient Competition (ITC) and Spatial Distance Stretching (SDS) in the optimization process. Specifically, we utilize a twin-network-like framework to generate both non-targeted and targeted AEs, introducing a new competition mechanism ITC where non-targeted adversarial gradients are applied each epoch to hinder the optimization of targeted adversarial perturbations, thus enhancing robustness in targeted attacks. Additionally, a top-k SDS strategy is employed, guiding AEs to penetrate target class regions in the latent multi-dimensional space while globally distancing from multiple closest non-targeted regions, ultimately achieving optimal adversarial transferability. Compared with state-of-the-art competition-based attacks, our method demonstrates significant transferability advantages, with average transferable TASRs improved by 16.1% and 21.4% on mainstream CNNs and ViTs, respectively, while also achieving an unmatched breaking-through defense capability.

World models allow autonomous agents to plan and explore by predicting the visual outcomes of different actions. However, for robot manipulation, it is challenging to accurately model the fine-grained robot-object interaction within the visual space using existing methods which overlooks precise alignment between each action and the corresponding frame. In this paper, we present IRASim, a novel world model capable of generating videos with fine-grained robotobject interaction details, conditioned on historical observations and robot action trajectories. We train a diffusion transformer and introduce a novel frame-level actionconditioning module within each transformer block to explicitly model and strengthen the action-frame alignment. Extensive experiments show that: (1) the quality of the videos generated by our method surpasses all the baseline methods and scales effectively with increased model size and computation; (2) policy evaluations using IRASim exhibit a strong correlation with those using the ground-truth simulator, highlighting its potential to accelerate real-world policy evaluation; (3) testing-time scaling through model-based planning with IRASim significantly enhances policy performance, as evidenced by an improvement in the IoU metric on the Push-T benchmark from 0.637 to 0.961; (4) IRASim provides flexible action controllability, allowing virtual robotic arms in datasets to be controlled via a keyboard or VR controller. Video and code are available at https://gen-irasim.github.io/.

Federated learning (FL) is becoming increasingly popular due to concerns about data privacy. In FL, multiple clients and a server jointly train a model. Clients update their models locally and send only the updated parameters to the server rather than raw data, which helps protect data privacy to a certain extent. However, research shows that even when the server or other clients know a client’s parameter data, adversaries can recover a lot of private information about data owners through inference attacks. In addition, adversaries can also obtain private information from aggregated parameters. To solve these problems, we propose a new privacy-preserving scheme in cross-silo FL based on aggregated multi-key BFV homomorphic encryption and noise addition. In this work, we introduce an aggregated multi-key BFV homomorphic encryption method to protect data privacy for each single transmission and a noise addition method to enhance privacy protection for aggregated parameters. Our scheme protects data privacy against collusion between the server and up to N − 2 clients without relying on a trusted third party, where N is the total number of clients participating in the training process. The experiments show that our noise-assisted aggregated multi-key BFV homomorphic encryption method achieves the same model performance as a plain federated averaging algorithm with the same machine learning models while providing higher computational and communication efficiency compared to two other state-of-the-art methods.

As large language models (LLMs) are increasingly deployed on resource-constrained edge devices, quantization techniques have been widely adopted to reduce model size and computational requirements. However, this process can expose models to new vulnerabilities. In this work, we introduce the Quantization Conditioned Misalignment (Q-Misalign) attack, a novel threat in which safety misalignment remains dormant in a full-precision LLM but becomes exploitable post-quantization. We demonstrate that our Q-Misalign attack effectively bypasses safety mechanisms and enables the generation of harmful content in quantized models while maintaining full-precision performance. Furthermore, we propose a contrastive task vector-based approach to enhance attack durability, ensuring that vulnerabilities persist even after downstream fine-tuning. Experimental results show that Q-Misalign attack significantly increases jailbreak success rates in quantized models, while preserving model utility and safety alignment in full precision. Our findings highlight a critical gap in current LLM safety measures and call for more robust defenses in quantization-aware scenarios.

Incorporating user preferences into large language models (LLMs) can enhance the personalization and reliability of model outputs and facilitate the application of LLMs to real-world scenarios. However, leveraging user preferences can be a double-edged sword. Recent studies have found that improper utilization can incur sycophancy, where LLMs prioritize alignment with user preferences over the correctness of their outputs. To address sycophancy in LLMs, we analyze and model the problem through the lens of structured causal models (SCMs). We attribute sycophancy to LLMs' reliance on spurious correlations between user preferences and model outputs in this paper. Based on the proposed SCMs, we develop a novel framework, termed **CAUSM**, to mitigate sycophancy in LLMs by exploiting a significant causal signature. Specifically, we eliminate the spurious correlations embedded in the intermediate layers of LLMs through causally motivated head reweighting, and then calibrate the intra-head knowledge along the causal representation direction. Extensive experiments are conducted across diverse language tasks to demonstrate the superiority of our method over state-of-the-art competitors in mitigating sycophancy in LLMs.

Large language models (LLMs) have been shown to face hallucination issues due to the data they trained on often containing human bias; whether this is reflected in the decision-making process of LLM agents remains under-explored. As LLM Agents are increasingly employed in intricate social environments, a pressing and natural question emerges: Can we utilize LLM Agents' systematic hallucinations to mirror human cognitive biases, thus exhibiting irrational social intelligence? In this paper, we probe the irrational behavior among contemporary LLM agents by melding practical social science experiments with theoretical insights. Specifically, we propose CogMir, an open-ended Multi-LLM Agents framework that utilizes hallucination properties to assess and enhance LLM Agents’ social intelligence through cognitive biases. Experimental results on CogMir subsets show that LLM Agents and humans exhibit high consistency in irrational and prosocial decision-making under uncertain conditions, underscoring the prosociality of LLM Agents as social entities and highlighting the significance of hallucination properties. Additionally, CogMir framework demonstrates its potential as a valuable platform for encouraging more research into the social intelligence of LLM Agents.

The next-coordinate prediction paradigm has emerged as the de facto standard in current auto-regressive mesh generation methods. Despite their effectiveness, there is no efficient measurement for the various tokenizers that serialize meshes into sequences. In this paper, we introduce a new metric Per-Token-Mesh-Entropy (PTME) to evaluate the existing mesh tokenizers theoretically without any training. Building upon PTME, we propose a plug-and-play tokenization technique called coordinate merging. It further improves the compression ratios of existing tokenizers by rearranging and merging the most frequent patterns of coordinates. Through experiments on various tokenization methods like MeshXL, MeshAnything V2, and Edgerunner, we further validate the performance of our method. We hope that the proposed PTME and coordinate merging can enhance the existing mesh tokenizers and guide the further development of native mesh generation.

Compositional Zero-Shot Learning (CZSL) aims to recognize novel state-object compositions by leveraging the shared knowledge of their primitive components. Despite considerable progress, effectively calibrating the bias between semantically similar multimodal representations, as well as generalizing pre-trained knowledge to novel compositional contexts, remains an enduring challenge. In this paper, our interest is to revisit the conditional transport (CT) theory and its homology to the visual-semantics interaction in CZSL and further, propose a novel Trisets Consistency Alignment framework (dubbed TsCA) that well-addresses these issues. Concretely, we utilize three distinct yet semantically homologous sets, i.e., patches, primitives, and compositions, to construct pairwise CT costs to minimize their semantic discrepancies. To further ensure the consistency transfer within these sets, we implement a cycle-consistency constraint that refines the learning by guaranteeing the feature consistency of the self-mapping during transport flow, regardless of modality. Moreover, we extend the CT plans to an open-world setting, which enables the model to effectively filter out unfeasible pairs, thereby speeding up the inference as well as increasing the accuracy. Extensive experiments are conducted to verify the effectiveness of the proposed method. The code is available at https://github.com/keepgoingjkg/TsCA.

The mixture of experts (MoE) model is a sparse variant of large language models (LLMs), designed to hold a better balance between intelligent capability and computational overhead. Despite its benefits, MoE is still too expensive to deploy on resource-constrained edge devices, especially with the demands of on-device inference services. Recent research efforts often apply model compression techniques, such as quantization, pruning and merging, to restrict MoE complexity. Unfortunately, due to their predefined static model optimization strategies, they cannot always achieve the desired quality-overhead trade-off when handling multiple requests, finally degrading the on-device quality of service. These limitations motivate us to propose the D2MoE, an algorithm-system co-design framework that matches diverse task requirements by dynamically allocating the most proper bit-width to each expert. Specifically, inspired by the nested structure of matryoshka dolls, we propose the matryoshka weight quantization (MWQ) to progressively compress expert weights in a bit-nested manner and reduce the required runtime memory. On top of it, we further optimize the I/O-computation pipeline and design a heuristic scheduling algorithm following our hottest-expert-bit-first (HEBF) principle, which maximizes the expert parallelism between I/O and computation queue under constrained memory budgets, thus significantly reducing the idle temporal bubbles waiting for the experts to load. Evaluations on real edge devices show that D2MoE improves the overall inference throughput by up to 1.39× and reduces the peak memory footprint by up to 53% over the latest on-device inference frameworks, while still preserving comparable serving accuracy as its INT8 counterparts.

Despite the significant breakthrough of Mixture-of-Experts (MoE), the increasing scale of these MoE models presents huge memory and storage challenges. Existing MoE pruning methods, which involve reducing parameter size with a uniform sparsity across all layers, often lead to suboptimal outcomes and performance degradation due to varying expert redundancy in different MoE layers. To address this, we propose a non-uniform pruning strategy, dubbed Differentiable Expert Pruning (DiEP), which adaptively adjusts pruning rates at the layer level while jointly learning inter-layer importance, effectively capturing the varying redundancy across different MoE layers. By transforming the global discrete search space into a continuous one, our method handles exponentially growing non-uniform expert combinations, enabling adaptive gradient-based pruning. Extensive experiments on five advanced MoE models demonstrate the efficacy of our method across various NLP tasks. Notably, \textbf{DiEP} retains around 92\% of original performance on Mixtral 8$\times$7B with only half the experts, outperforming other pruning methods by up to 7.1% on the challenging MMLU dataset.

Existing pretrained models for 3D mesh generation often suffer from data biases and produce low-quality results, while global reinforcement learning (RL) methods rely on object-level rewards that struggle to capture local structure details. To address these challenges, we present $\textbf{Mesh-RFT}$, a novel fine-grained reinforcement fine-tuning framework that employs Masked Direct Preference Optimization (M-DPO) to enable localized refinement via quality-aware face masking. To facilitate efficient quality evaluation, we introduce an objective topology-aware scoring system to evaluate geometric integrity and topological regularity at both object and face levels through two metrics: Boundary Edge Ratio (BER) and Topology Score (TS). By integrating these metrics into a fine-grained RL strategy, Mesh-RFT becomes the first method to optimize mesh quality at the granularity of individual faces, resolving localized errors while preserving global coherence. Experiment results show that our M-DPO approach reduces Hausdorff Distance (HD) by 24.6\% and improves Topology Score (TS) by 3.8\% over pre-trained models, while outperforming global DPO methods with a 17.4\% HD reduction and 4.9\% TS gain. These results demonstrate Mesh-RFT’s ability to improve geometric integrity and topological regularity, achieving new state-of-the-art performance in production-ready mesh generation.

The widespread deployment of large language models (LLMs) allows users to access their capabilities via black-box APIs, but backdoor attacks pose serious security risks for API users by hijacking the model behavior. This highlights the importance of backdoor detection technologies to help users audit LLMs before use. However, most existing LLM backdoor defenses require white-box access or costly reverse engineering, limiting their practicality for resource-constrained users. Moreover, they mainly target classification tasks, leaving broader generative scenarios underexplored. To solve the problem, this paper introduces ICLScan, a lightweight framework that exploits targeted in-context learning (ICL) as illumination for backdoor detection in black-box LLMs, which effectively supports generative tasks without additional training or model modifications. ICLScan is based on our finding of backdoor susceptibility amplification: LLMs with pre-embedded backdoors are highly susceptible to new trigger implantation via ICL. Including only a small ratio of backdoor examples (containing ICL-triggered input and target output) in the ICL prompt can induce ICL trigger-specific malicious behavior in backdoored LLMs. ICLScan leverages this phenomenon to detect backdoored LLMs by statistically analyzing whether the success rate of new trigger injection via targeted ICL exceeds a threshold. It requires only multiple queries to estimate the backdoor success rate, overcoming black-box access and computational resource limitations. Extensive experiments across diverse LLMs and backdoor attacks demonstrate ICLScan's effectiveness and efficiency, achieving near-perfect detection performance (precision/recall/F1-score/ROC-AUC all approaching 1) with minimal additional overhead across all settings.

The Metaverse, with its promise of immersive experiences and transformative user interactions, represents a new paradigm for IoT development. By integrating IoT with the Metaverse, real-world data can seamlessly enrich virtual environments, offering diverse choices to users. However, the sheer volume of products and user groups in the Metaverse poses challenges in effectively matching users with suitable products. Recommendation systems, particularly Collaborative Filtering (CF), emerge as a solution to this issue, leveraging user preferences and social dynamics. However, traditional CF algorithms face efficiency challenges in the multi-dimensional data landscape of the Metaverse. To address this, a clustering-based CF algorithm is proposed, enhancing recommendation efficiency by leveraging social connections. Additionally, the recommendation system is enhanced with a DSD-Transformer framework, optimizing recommendation accuracy. The experiments indicate that our proposed method may considerably enhance the Metaverse experience when compare to various sophisticated methods and can be utilized to build a range of product recommendation systems.

Federated learning (FL), as an innovative paradigm of distributed learning, provides reliable support for the growing edge intelligence (EI). The limitations of traditional FL’s reliance on ground base stations (BSs) make the development of aerial server unmanned aerial vehicles (UAVs) inevitable, thereby developing the air-ground integrated FL (AGIFL). However, current efforts predominately focus on static offline training based on existing datasets and some efforts consider online training in dynamic sample environments, where new samples need to be fully pre-trained to determine sample quality. To this end, we study how to realize high-performance of FL in dynamic environment without training all samples. Specifically, we formulate a joint optimization problem for sample selection, UAV deployment, and resource distribution aiming to minimize the trade-off between the user energy consumption and FL performance. To address the optimization problem without explicit expression, we employ meta-learning to derive an upper bound on the gradient norm of the loss function to evaluate learning performance, and describe how time-varying small-batch ratios affect this bound. Then, we propose an optimization algorithm that ensures convergence, capitalizing on the block coordinate descent techniques. To demonstrate the efficacy of our algorithm, we conduct both extensive simulations and proof-of-concept field experiments. The findings indicate an average improvement of approximately 39% in reducing the objective value when compared to the benchmarks.

The Metaverse is a user-generated virtual world, aiming to provide highly personalized experiences for users. A product personalization design platform is a critical direction for the Metaverse’s future development, enhancing user experience by offering personalized services. Blockchain technology ensures the security and privacy of user data, and enables personalized services through smart contracts, offering opportunities for personalization platforms. However, blockchain’s decentralization can lead to excessive product data, resulting in ineffective data management and optimization, subsequently confusing personalized product design styles, which diminishes user experience. To address these issues, this study proposes the product style automatic generation system (PSAGS), centered on an image generation unit. The system outputs images with style information based on the input product text, achieving precise quantization and visualization of product styles, thereby enhancing user engagement and loyalty to the Metaverse. The image generation unit, with a standardization module can standardize the product style, namely, the relationship between product design elements and user emotions, addressing the problem of managing vast style data due to blockchain’s decentralization. The generation module utilizes a diffusion model enhanced with contrastive language-image pretraining (CLIP) to generate style images, deepening the Metaverse experience. Optimizations include dilation convolution in the UNet architecture to enhance image quality and fine-grained CLIP transformations for improved image and text alignment. Results demonstrate the system’s effectiveness in streamlining design processes and improving image quality in personalized product design, with wide applications in the Metaverse.

With the rapid development of on-board computing technology, on-orbit information processing has become a new direction for reducing service response delays and improving the quality of space-based information services. Especially in space-air–ground integrated applications in 6G networks, remote sensing image processing tasks are highly important because of their critical role in applications such as environmental monitoring and public safety. However, the fluctuations in data volume due to significant scene differences, along with the limitations in individual satellite capabilities caused by size and power constraints, present new challenges for on-orbit image processing. To address these challenges, we model a data-driven on-orbit resource scheduling problem for space-air-ground integrated networks based on distributionally robust optimization, aiming to minimize the average image processing delay. We first construct an ambiguity set based on the Wasserstein distance and the historical distribution of image data, which helps transform the original upper-bound expectation problem into an explicitly expressed mixed-integer nonlinear (MINLP) problem. Furthermore, to reduce complexity and expedite the solution process, we decouple the MINLP problem into three subproblems using the block coordinate descent method and designed an iterative solving algorithm. The numerical results demonstrate that our proposed method achieves better fitting accuracy than traditional methods and reduces the average image processing delay.

Decentralized Knowledge Graph (DKG) platforms allow the sharing of knowledge with multiple owners. While data owners can share their data with others by encrypting their data before sharing it, this naïve approach prevents data encrypted by different owners from being queried together, as it compromises query verifiability, an essential DKG platform feature. We propose Pistis, the first DKG platform capable of preserving ownership while also enabling verifiable SPARQL queries. Two novel techniques facilitate this: owner-managed end-to-end encryption and collaborative query verification. In Pistis, data owners thus encrypt their data individually and collaborate to construct an authenticated data structure (ADS) with a global key by means of secret sharing and secure multi-party computation. Then, by indexing KG data as ciphertext over the ADS, Pistis offers a cryptographic scheme called VO-SPARQL that facilitates verifiable queries on encrypted KG data with multiple owners. Pistis provides succinct proofs for two-stage SPARQL queries, including subgraph queries based on the ADS and aggregation on encrypted intermediate results based on a key-aggregate cryptographic primitive. A theoretical analysis and an empirical study provide detailed insight into the performance of Pistis while offering provable security.

Parallelizing CNN inference on heterogeneous edge clusters with data parallelism has gained popularity as a way to meet real-time requirements without sacrificing model accuracy. However, existing algorithms struggle to find optimal parallel granularity for complex CNNS, the structure of which is a directed acyclic graph (DAG) rather than a chain, and the parallel dimension is inflexible. To distribute the workload of modern CNNs on heterogeneous devices is also proven as NP-hard problem. In this article, we introduce DeepZoning , a versatile and cooperative inference framework that combines both model and data parallelism to accelerate CNN inference. DeepZoning employs two algorithms at different levels: (1) a low-level Adaptive Workload Partition algorithm that uses linear programming and takes spatial and channel dimensions into optimization during the search for feature map distribution on heterogeneous devices, and (2) a high-level Model Partition algorithm that finds the optimal model granularity and organizes complex CNNs into sequential zones to balance communication and computation during execution. Our experimental evaluations show that DeepZoning is effective, achieving up to a 3.02× speed improvement on our experimental prototype compared to state-of-the-art algorithms.

Federated learning (FL) enables privacy-preserving distributed machine learning by training models on edge client devices using their local data without revealing their raw data. In edge environments, various applications require different neural network models, making it crucial to perform joint training of multiple models on edge devices, known as multi-task FL. While existing multi-task FL approaches enhance resource utilization on edge devices through adaptive resource configuration or client selection, optimizing either of these aspects alone may lead to suboptimality. Therefore, in this paper, we explore a joint client selection and resource configuration method called JCSRC for multi-task FL, aiming to maximize energy efficiency in environments with limited computation and communication resources and heterogeneous client devices. Firstly, we formalize this problem as a mixed-integer nonlinear programming problem considering all these characteristics and prove its NP-hardness. To address this problem, we first design a multi-agent reinforcement learning (MARL)-based client selection method that selects appropriate clients for each task to train their models. The MARL method makes client selection decisions based on the clients’ data quality, energy efficiency, communication, and computation capacity to ensure fast convergence and energy efficiency. Then, we design a particle swarm optimization (PSO)-based resource configuration scheme that configures appropriate computation and bandwidth resources for each task on each client. The PSO scheme makes resource configuration decisions based on theoretically derived optimal CPU frequency and bandwidth to achieve high energy efficiency. Finally, we carry out extensive simulations and testbed-based experiments to validate our proposed JCSRC. The results demonstrate that, in comparison to state-of-the-art solutions, JCSRC can save energy consumption by up to 59% to achieve the target accuracy.

Personalized federated learning (pFL) is a promising approach to train customized models for multiple clients over heterogeneous data distributions. However, existing works on pFL often rely on the optimization of model parameters and ignore the personalization demand on neural network architecture, which can greatly affect the model performance in practice. Therefore, generating personalized models with different neural architectures for different clients is a key issue in implementing pFL in a heterogeneous environment. Motivated by Neural Architecture Search (NAS), a model architecture searching methodology, this paper aims to automate the model design in a collaborative manner while achieving good training performance for each client. Specifically, we reconstruct the centralized searching of NAS into the distributed scheme called Personalized Architecture Search (PAS), where differentiable architecture fine-tuning is achieved via gradient-descent optimization, thus making each client obtain the most appropriate model. Furthermore, to aggregate knowledge from heterogeneous neural architectures, a knowledge distillation-based training framework is proposed to achieve a good trade-off between generalization and personalization in federated learning. Extensive experiments demonstrate that our architecture-level personalization method achieves higher accuracy under the non-iid settings, while not aggravating model complexity over state-of-the-art benchmarks.

Federated Learning (FL) has emerged as a promising technique for collaboratively training machine learning models among multiple participants while preserving privacy-sensitive data. However, the conventional parameter server architecture presents challenges in terms of communication overhead when employing iterative optimization methods such as Stochastic Gradient Descent (SGD). Although communication compression techniques can reduce the traffic cost of FL during each training round, they often lead to degraded convergence rates, mainly due to compression errors and data heterogeneity. To address these issues, this paper presents FedQClip, an innovative approach that combines quantization and Clipped SGD. FedQClip leverages an adaptive step size inversely proportional to the$\ell_{2}$norm of the gradient, effectively mitigating the negative impacts of quantized errors. Additionally, clipped operations can be applied locally and globally to further expedite training. Theoretical analyses provide evidence that, even under the settings of Non-IID (non-independent and identically distributed) data, FedQClip achieves a convergence rate of$\mathcal{O}(\frac{1}{\sqrt{T}})$, effectively addressing the convergence degradation caused by compression errors. Furthermore, our theoretical analysis highlights the importance of selecting an appropriate number of local updates to enhance the convergence of FL training. Through extensive experiments, we demonstrate that FedQClip outperforms state-of-the-art methods in terms of communication efficiency and convergence rate.

Federated Learning with Cloud-Edge Collaboration (FL-CEC) has emerged as a cutting-edge paradigm in distributed learning. Efficient resource investment incentive mechanisms are crucial to encouraging clients in FL-CEC to contribute the necessary data and computational resources for training. However, existing studies are inadequate in meeting the incentive design requirements under multi-level information-sharing scenarios. Moreover, current works often rely on specific functional relationships between resource investment and global model accuracy. To bridge these gaps, this paper investigates the incentive problem for data and computational resource investment under multi-level information-sharing levels. We design a resource investment incentive mechanism based on a weighted potential game without depending on any specific functional relationship between data investment and model accuracy. Furthermore, we propose four algorithms to solve resource investment strategies for different levels of information sharing. The complexity and convergence rates of the proposed algorithms are thoroughly analyzed. Finally, we construct a simulation incentive platform on the Aliyun. Extensive evaluations demonstrate that the proposed scheme effectively enhances social welfare, and improves collaborative training accuracy and efficiency.

Cloud-Edge Collaborative Architecture (CEA) is a prominent framework that provides low-latency and energy-efficient solutions for video stream processing. In Cloud-Edge Collaborative Video Streaming Systems (CEAVS), efficient online offloading strategies for video tasks are crucial for enhancing user experience. However, most existing works overlook budget constraints, which limits their applicability in real-world scenarios constrained by finite resources. Moreover, they fail to adequately address the heterogeneity of video task redundancies, leading to suboptimal utilization of CEAVS's limited resources. To bridge these gaps, we propose an Efficient Online Computing framework for CEAVS (EOCA) that jointly optimizes accuracy, energy consumption, and latency performance through adaptive online offloading and redundancy compression, without requiring future task information. Technically, we formulate computing offloading and adaptive compression under budget constraints as a stochastic optimization problem that maximizes system satisfaction, defined as a weighted combination of accuracy, latency, and energy performance. We employ Lyapunov optimization to decouple the long-term budget constraint. We prove that the decoupled problem is a generalized ordinal potential game and propose algorithms based on generalized Benders decomposition (GBD) and the best response to obtain Nash equilibrium strategies for computing offloading and task compression. Finally, we analyze EOCA's performance bound, convergence rate, and worst-case performance guarantees. Evaluations demonstrate that EOCA effectively improves satisfaction while effectively balancing satisfaction and computational overhead.

Prediction queries have revolutionized data search by integrating machine learning models and traditional data processing operations for advanced analytics. However, existing prediction query frameworks for outsourced databases face a critical security vulnerability: data flows are processed in plaintext on semi-honest servers, making them susceptible to data breaches. The main challenge in achieving secure prediction queries is that machine learning inference and data processing operations are distinct functionalities, while most current cryptographic frameworks support only a single type of operation on specific encrypted data. To bridge this crucial gap, we propose$\mathsf {SecPQ}$, the first framework tailored for secure prediction queries. Our approach unifies decision tree pipelines and data processing operations, such as selection, projection, and equality-joining, through equality matching on encrypted outsourced data. This enables the design of secure prediction queries with decision tree pipelines operating on encrypted data. We provide formal security definitions and proofs for$\mathsf {SecPQ}$. To further optimize the efficiency of secure prediction queries, we leverage order-preserving encryption to construct$\mathsf {SecPQ}_{{{ope}}}$, which offers improved query efficiency at the expense of weaker security properties compared with$\mathsf {SecPQ}$. Extensive experimental evaluations on billions of records demonstrate the feasibility and effectiveness of both$\mathsf {SecPQ}$and$\mathsf {SecPQ}_{{{ope}}}$.

Federated learning (FL) has attracted tremendous attention due to its capability to preserve data privacy. In FL, a parameter server (PS) without accessing clients' raw data can assist decentralized clients in completing model training by aggregating and distributing model parameters for multiple iterations. From the clients' perspective, exposing model parameters to the PS can still result in privacy leakage. To further enhance privacy protection, differentially private federated learning (DPFL) is invented, in which clients add differentially private (DP) noises to distort their parameters to be exposed. However, the main challenge of DPFL lies in inferior model accuracy due to DP noises. To overcome this challenge, in this paper we propose a novel algorithmic framework for DPFL, which is called CODP, by cascading and offsetting DP noises between iterations. In existing works, each DPFL client only considers how to protect its model parameters based on the number of iterations to expose parameters overlooking the underlying relation of model parameters in consecutive iterations. The novelty of CODP lies in cascading DP noises from each iteration to its subsequent iteration so that DP noises can be offset in the subsequent iteration, and hence model accuracy can be improved. Additionally, we theoretically prove that CODP can substantially improve the convergence rate of DPFL without compromising privacy preservation by leveraging the most widely used Laplace and Gaussian mechanisms, respectively. We conduct comprehensive experiments using MNIST, Fashion-MNIST, and Lending Club datasets to demonstrate that the model accuracy of DPFL can be remarkably improved by CODP with a fixed privacy budget.

Machine learning-based Android malware detection has consistently demonstrated superior results. However, with the continual evolution of the Android framework, the efficacy of the deployed models declines markedly. Existing solutions necessitate frequent and expensive model retraining to resist the constant evolution of malware accompanying Android framework updates. To address this, we introduce a solution called ASDroid, which generalizes specific APIs into similar API clusters to counteract evolving Android malware threats. One primary challenge lies in identifying analogous API clusters that correspond to specific APIs. Our approach involves extracting semantic information from open-source API source code to construct a heterogeneous information graph, and utilizing embedding algorithms to obtain semantic vector representations of APIs. APIs that are close in embedding distance are presumed to have similar semantics. Our dataset encompasses Android applications spanning nine years from 2011 to 2019. In comparison to existing Android malware detection model aging mitigation solutions like APIGraph, SDAC and MaMaDroid, ASDroid demonstrates greater accuracy and more effective at resisting continuously evolving malware.

Age of Information (AoI) has been proposed as a new performance metric to capture the freshness of data. At wireless-powered network edge, the source nodes first need to be charged ready for update transmissions, which means the system AoI is not only decided by the scheduling of update transmissions but also by the designing of charging plan. However, the existing works either only focused on the point of scheduling update transmissions or have a rigid assumption that only one source node can be charged per time. Aiming at making the work more practical and general, we investigate the average AoI optimization problem at wireless-powered network edge with directional charging. Firstly, the theoretical bound of the weighted sum of average AoI of the entire network with a directional charger is analyzed, which is proved to be related to nodes' maximum transmitting interval and the charging strategy. An optimal charging time computation algorithm is proposed to obtain the maximum transmitting interval of each source node by considering the overlapped areas of different charging orientations. After then, an AoI-aware periodical charging scheduling algorithm is proposed to compute a periodical charging schedule while the average AoI is bounded, including a charging period$T$and the charging orientations assigned to each time slot within$T$. The proposed algorithm is proved to have an approximation ratio of up to 1.5625. Furthermore, several approximate algorithms are also proposed for average AoI optimization with multiple chargers and network bandwidth constraint. Finally, the extensive simulations demonstrate the high performance of the proposed algorithm in terms of AoI.

Generative diffusion models like Stable Diffusion are at the forefront of the thriving field of generative models today, celebrated for their robust training methodologies and high-quality photorealistic generation capabilities. These models excel in producing rich content, establishing them as essential tools in the industry. Building on this foundation, the field has seen the rise ofpersonalized content synthesisas a particularly exciting application. However, the large model sizes and iterative nature of inference make it difficult to deploy personalized diffusion models broadly on local devices with heterogeneous computational power. To address this, we propose a novel framework for efficient multi-user offloading of personalized diffusion models. This framework accommodates a variable number of users, each with different computational capabilities, and adapts to the fluctuating computational resources available on edge servers. To enhance computational efficiency and alleviate the storage burden on edge servers, we propose a tailored multi-user hybrid inference approach. This method splits the inference process for each user into two phases, with an optimizable split point. Initially, a cluster-wide model processes low-level semantic information for each user's prompt using batching techniques. Subsequently, users employ their personalized models to refine these details during the later phase of inference. Given the constraints on edge server computational resources and users' preferences for low latency and high accuracy, we model the joint optimization of each user's offloading request handling and split point as an extension of the Generalized Quadratic Assignment Problem (GQAP). Our objective is to maximize a comprehensive metric that balances both latency and accuracy across all users. To solve this NP-hard problem, we transform the GQAP into an adaptive decision sequence, model it as a Markov decision process, and develop a hybrid solution combining deep reinforcement learning with convex optimization techniques. Simulation results validate the effectiveness of our framework, demonstrating superior optimality and low complexity compared to traditional methods. All related code, datasets, and fine-tuned models are available athttps://github.com/wty2011jl/E-MOPDM.

Radio frequency identification (RFID) system has been extensively employed to track missing items by affixing them with RFID tags. Many practical applications require to efficiently identify missing events for a specific subset of system tags (called key tags) due to their elevated importance. Existing methods primarily aim to identify all tags, which makes it challenging to specifically identify key tags because of interference from other non-key tags (called ordinary tags). In light of this, several key tag identification methods follow a two-step scheme that filters ordinary tags first and then identifies key tags. Nevertheless, this wastes too much time on tag filtering, resulting in low time efficiency. This paper presents a novel missing key tag identification protocol with two creative designs to gain high efficiency. First, we develop a novel verification technique that can rapidly determine the presence or absence of key tags amid the scenarios with both key tags and ordinary ones. By combining the ON-OFF Keying modulation, we could verify multiple key tags in a single slot, thereby reducing the total slots required. Second, we design a new selection technique that efficiently selects the unverified key tags for further verification, while filtering out the verified key tags and irrelevant ordinary tags to avoid redundant data transmission. Additionally, we present an enhancement protocol that leverages a preselection technique to avoid collecting useless tag responses, further boosting efficiency. We carry out rigorous theoretical analysis to optimize the performance of the proposed protocols. Both simulations and practical experiments demonstrate that our method is markedly superior to state-of-the-art solutions.

The edge intelligence (EI) has been widely applied recently. Splitting the model between device, edge server, and cloud can significantly improve the performance of EI. The model segmentation without user mobility has been investigated in detail in previous studies. However, in most EI use cases, the end devices are mobile. Few studies have been conducted on this topic. These works still have many issues, such as ignoring the energy consumption of mobile device, inappropriate network assumption, and low effectiveness on adapting user mobility, etc. Therefore, to address the disadvantages of model segmentation and resource allocation in previous studies, we propose mobility and cost aware model segmentation and resource allocation algorithm for accelerating the inference at edge (MCSA). Specifically, in the scenario without user mobility, the loop iteration gradient descent (Li-GD) algorithm is provided. When the mobile user has a large model inference task that needs to be calculated, it will take the energy consumption of mobile user, the communication and computing resource renting cost, and the inference delay into account to find the optimal model segmentation and resource allocation strategy. In the scenario with user mobility, the mobility aware Li-GD (MLi-GD) algorithm is proposed to calculate the optimal strategy. Then, the properties of the proposed algorithms are investigated, including convergence, complexity, and approximation ratio. The experimental results demonstrate the effectiveness of the proposed algorithms.

Edge computing significantly enhanced the growth of edge-assistant video streaming applications. However, challenges such as unpredictable wireless conditions, resource constraints, and task redundancy have intertwined impacts on the overall performance of edge video streaming systems (EVS). Therefore, it is essential to have an integrated framework that addresses resource management, computational offloading, and video task preprocessing. Existing optimization strategies often neglect the simultaneous management of computational offloading, resource allocation, and video task preprocessing, leading to a suboptimal system utility. Moreover, they struggle to handle high-dimensional decision variables. On the other hand, learning-based adaptive schemes fall short in integrating distributed decisions and ensuring the scalability of wireless devices. Additionally, current approaches lack adaptive incentives. To bridge these gaps, we propose a novel framework called AIRA, which is based on improved multi-agent reinforcement learning (MARL) and smart contracts. AIRA manages resources, video compression, and adaptive incentives in a distributed manner. It consists of a MARL-driven cooperative learning algorithm (CLA) and a smart contract-guided adaptive incentive mechanism. Leveraging an actor-critic structure, the CLA enables wireless devices to master strategies for resource allocation, video task compression, and offloading, utilizing historical data. Notably, the CLA incorporates an attention mechanism to select pivotal tuples from the observation-action pairings among different agents, ensuring improved scalability and computational prowess. Evaluations based on real-world trajectories demonstrate that AIRA enables adaptive incentives. Compared to state-of-the-art approaches, CLA effectively enhances the long-term system utility and scalability of EVS.

The prevalence of the pre-training & fine-tuning paradigm enables machine learning models to quickly adapt to various downstream tasks by fine-tuning pre-trained foundation models (FMs), greatly facilitating various IoT applications that rely on model inference in dynamic edge serverless environments. Efficiently dispatching inference requests and configuring instances to batch inference requests can significantly enhance resource efficiency. However, existing serverless inference solutions are tailored for traditional models, make coarse-grained request dispatching and instance configuration decisions, fail to exploit the shared model backbone characteristics of the FM and capture delayed rewards in dynamic environments, and ignore communication latency between edge sites, resulting in high costs and constraint violations. In this paper, we leverage our insight that fine-grained batch inference requests can effectively exploit the shared model backbone feature of FM to save monetary costs. We propose an algorithm that incorporates deep reinforcement learning (DRL) and expert intervention for fine-grained request dispatching and instance configuration, where the DRL component outputs fractional solutions as guidance, while the expert intervention module integrates our insights—batching reduces monetary costs at the expense of increased inference latency, whereas higher configurations shorten inference latency. This module rounds fractional solutions and adjusts instance configurations to search for optimal solutions while satisfying constraints, with theoretical guarantees rigorously proved. Finally, we conducted our experiments on an OpenFaas-based platform and simulator, and extensive trace-driven evaluation results show that ExpertDRL can save costs by up to 85.14% and improve request acceptance ratio by up to 26.93%, compared to the state-of-the-art solution.

Personalized federated learning (pFL) is to collaboratively train non-identical machine learning models for different clients to adapt to their heterogeneously distributed datasets. State-of-the-art pFL approaches pay much attention on exploiting clients’ inter-similarities to facilitate the collaborative learning process, meanwhile, can barely escape from the irrelevant knowledge pooling that is inevitable during the aggregation phase, and thus hindering the optimization convergence and degrading the personalization performance. To tackle such conflicts between facilitating collaboration and promoting personalization, we propose a novel pFL framework, dubbed pFedC, to first decompose the global aggregated knowledge into several compositional branches, and then selectively reassemble the relevant branches for supporting conflicts-aware collaboration among contradictory clients. Specifically, by reconstructing each local model into a shared feature extractor and multiple decomposed task-specific classifiers, the training on each client transforms into a mutually reinforced and relatively independent multi-task learning process, which provides a new perspective for pFL. Besides, we conduct a purified knowledge aggregation mechanism via quantifying the combination weights for each client to capture clients’ common prior, as well as mitigate potential conflicts from the divergent knowledge caused by the heterogeneous data. Extensive experiments over various models and datasets demonstrate the effectiveness and superior performance of the proposed algorithm.

Foundation models (FMs) such as large language models are becoming the backbone technology for artificial intelligence systems. It is particularly challenging to deploy multiple FMs on edge devices, which not only have limited computational resources, but also encounter unseen input data from evolving domains or learning tasks. When new data arrives, existing prior art of FM mainly focuses on retraining compressed models of predetermined network architectures, limiting the feasibility of edge devices to efficiently achieve high accuracy for FMs. In this paper, we propose EdgeTA, a neuron-grained FM scaling system to maximize the overall accuracy of FMs promptly in response to their data dynamics. EdgeTA's key design features in scaling are (i) proxy mechanism, which adaptively transforms a FM into a compact architecture retaining the most important neurons to the input data, and (ii) neuron-grained scheduler, which jointly optimizes model sizes and resource allocation for all FMs on edge devices. Under tight retraining window and limited device resources, the design of EdgeTA can achieve most of the original FM's accuracy with much smaller retraining costs. We implement EdgeTA on FMs of natural language processing, computer vision and multimodal applications. Comparison results against state-of-the-art techniques show that our approach improves accuracy by 21.88% and reduces memory footprint and energy consumptions by 27.14% and 65.65%, while further achieving 15.96% overall accuracy improvement via neuron-grained scheduling.

Visual Simultaneous Localization and Mapping (vSLAM) is a prevailing technology for many emerging robotic applications. Achieving real-time SLAM on mobile robotic systems with limited computational resources is challenging because the complexity of SLAM algorithms increases over time. This restriction can be lifted by offloading computations to edge servers, forming the emerging paradigm ofedge-assisted SLAM. Nevertheless, the exogenous and stochastic input processes affect the dynamics of the edge-assisted SLAM system. Moreover, the requirements of clients on SLAM metrics change over time, exerting implicit and time-varying effects on the system. In this paper, we aim to push the limit beyond existing edge-assist SLAM by proposing a new architecture that can handle the input-driven processes and also satisfy clients’ implicit and time-varying requirements. The key innovations of our work involve a regional feature prediction method for importance-aware local data processing, a configuration adaptation policy that integrates data compression/decompression and task offloading, and an input-dependent learning framework for task scheduling with constraint satisfaction. Extensive experiments prove that our architecture improves pose estimation accuracy and saves up to 47% of communication costs compared with a popular edge-assisted SLAM system, as well as effectively satisfies the clients’ requirements.

Malware open-set recognition (MOSR) is an emerging research domain that aims at jointly classifying malware samples from known families and detecting the ones from novel unknown families, respectively. Existing works mostly rely on a well-trained classifier considering the predicted probabilities of each known family with a threshold-based detection to achieve the MOSR. However, our observation reveals that the feature distributions of malware samples are extremely similar to each other even between known and unknown families. Thus, the obtained classifier may produce overly high probabilities of testing unknown samples toward known families and degrade the model performance. In this article, we propose the multi\modal dual-embedding networks, dubbed MDENet, to take advantage of comprehensive malware features from different modalities to enhance the diversity of malware feature space, which is more representative and discriminative for down-stream recognition. Concretely, we first generate a malware image for each observed sample based on their numeric features using our proposed numeric encoder with a re- designed multiscale CNN structure, which can better explore their statistical and spatial correlations. Besides, we propose to organize tokenized malware features into a sentence for each sample considering its behaviors and dynamics, and utilize language models as the textual encoder to transform it into a representable and computable textual vector. Such parallel multimodal encoders can fuse the above two components to enhance the feature diversity. Last, to further guarantee the open-set recognition (OSR), we dually embed the fused multimodal representation into one primary space and an associated sub-space, i.e., discriminative and exclusive spaces, with contrastive sampling and -bounded enclosing sphere regularizations, which resort to classification and detection, respectively. Moreover, we also enrich our previously proposed large-scaled malware dataset MAL-100 with multimodal characteristics and contribute an improved version dubbed MAL-100+. Experimental results on the widely used malware dataset Mailing and the proposed MAL-100+ demonstrate the effectiveness of our method.

Extensive attention has been paid to the application of self-supervised learning (SSL) approaches on federated learning (FL) to tackle the label scarcity problem. Previous works on federated SSL (FedSSL) generally fall into two categories: parameter-based model aggregation or data-based feature sharing to achieve knowledge transfer among multiple unlabeled clients. Despite the progress, they inevitably rely on some assumptions, such as homogeneous models or the existence of an additional public dataset, which hinder the universality of the training frameworks for more general scenarios (e.g., unlabeled clients with heterogeneous models). Therefore, in this article, we propose a novel and general method named federated self-supervised learning with feature-correlation-based aggregation (FedFoA) to tackle the above limitations. By exchanging feature correlation instead of model parameters or feature mappings, our approach reduces the discrepancies of local representations learning processes, thus promoting collaboration between heterogeneous clients. A factorization-based method is designed to extract the cross-feature relation matrix from local representations, which serves as a knowledge medium for the aggregation phase. We demonstrate that FedFoA is a heterogeneity-supportive and privacy-preserving training framework and can be easily compatible with state-of-the-art FedSSL methods. Extensive empirical experiments demonstrate our proposed approach outperforms the state-of-the-art methods by a significant margin.

This article provides a novel framework for single-domain generalized object detection (i.e., Single-DGOD), where we are interested in learning and maintaining the semantic structures of self-augmented compound cross-domain samples to enhance the model's generalization ability. Different from domain generalized object detection (DGOD) trained on multiple source domains, Single-DGOD is far more challenging to generalize well to multiple target domains with only one single source domain. Existing methods mostly adopt a similar treatment from DGOD to learn domain-invariant features by decoupling or compressing the semantic space. However, there may exist two potential limitations: 1) pseudo attribute-label correlation due to extremely scarce single-domain data and 2) the semantic structural information is usually ignored, i.e., we found the affinities of instance-level semantic relations in samples are crucial to model generalization. In this article, we introduce semantic reasoning with compound domains (SRCD) for Single-DGOD. Specifically, our SRCD contains two main components, namely, the texture-based self-augmentation (TBSA) module and the local-global semantic reasoning (LGSR) module. TBSA aims to eliminate the effects of irrelevant attributes associated with labels, such as light, shadow, and color, at the image level by a light-yet-efficient self-augmentation. Moreover, LGSR is used to further model the semantic relationships on instance features to uncover and maintain the intrinsic semantic structures. Extensive experiments on multiple benchmarks demonstrate the effectiveness of the proposed SRCD. Code is available at github.com/zjrao/SRCD.

Mixed-precision quantization mostly predetermines the model bit-width settings before actual training due to the non-differential bit-width sampling process, obtaining suboptimal performance. Worse still, the conventional static quality-consistent training setting, i.e., all data is assumed to be of the same quality across training and inference, overlooks data quality changes in real-world applications which may lead to poor robustness of the quantized models. In this article, we propose a novel data quality-aware mixed-precision quantization framework, dubbed DQMQ, to dynamically adapt quantization bit-widths to different data qualities. The adaption is based on a bit-width decision policy that can be learned jointly with the quantization training. Concretely, DQMQ is modeled as a hybrid reinforcement learning (RL) task that combines model-based policy optimization with supervised quantization training. By relaxing the discrete bit-width sampling to a continuous probability distribution that is encoded with few learnable parameters, DQMQ is differentiable and can be directly optimized end-to-end with a hybrid optimization target considering both task performance and quantization benefits. Trained on mixed-quality image datasets, DQMQ can implicitly select the most proper bit-width for each layer when facing uneven input qualities. Extensive experiments on various benchmark datasets and networks demonstrate the superiority of DQMQ against existing fixed/mixed-precision quantization methods.

Mixture-of-Experts (MoE) is an emerging technique for scaling large models with sparse activation. MoE models are typically trained in a distributed manner with anexpert parallelismscheme, where experts in each MoE layer are distributed across multiple GPUs. However, the default expert parallelism suffers from the heavy network burden due to the all-to-all intermediate data exchange among GPUs before and after the expert run. Some existing works have proposed to reduce intermediate data exchanges by transferring experts to reduce the network loads, however, which would decrease parallelism level of expert execution and make computation inefficient. The weaknesses of existing works motivate us to explore whether it is possible to reduce inter-GPU traffic while maintaining a high degree of expert parallelism. This paper gives a positive response by presentingLuffy, a communication-efficient distributed MoE training system with two new techniques. First,Luffymigrates sequences among GPUs to hide heavy token pulling paths within GPUs and avoid copying experts over GPUs. Second, we propose token condensation that identifies similar tokens and then eliminates redundant transmissions. We implementLuffybased on PyTorch and evaluate its performance on a testbed of 16 V100 GPUs.Luffysystem can achieve a speedup of up to$2.73\times $compared to state-of-the-art MoE training systems.

State-of-the-art blockchain sharding solutions, such as Monoxide, can cause severely imbalanced distribution of transaction (TX) workloads across all blockchain shards due to the deployment policy of their accounts. Imbalanced TX distributions then producehot shards, in which the cross-shard TXs may experience an unlimited confirmation latency. Thus, how to address the hot-shard issue and how to reduce cross-shard TXs become significant challenges of blockchain sharding. Through reviewing the related studies, we find that a cross-shard TX protocol that can achieve workload balance among all shards and simultaneously reduce the quantity of cross-shard TXs is still absent from the literature. To this end, we propose BrokerChain, which is a cross-shard blockchain protocol dedicated to account-based state sharding. Essentially, BrokerChain exploits fine-grained state partition and account segmentation. We also elaborate on how BrokerChain handles cross-shard TXs through broker accounts. The security issues and other properties of BrokerChain are analyzed rigorously. Finally, we conduct comprehensive evaluations using an open-source blockchain sharding prototype namedBlockEmulator. The evaluation results show that BrokerChain outperforms other baselines in terms of transaction throughput, transaction confirmation latency, the queue size of the transaction pool, and workload balance.

Wireless Power Transfer technology, which can charge IoT devices over the air, has become a promising technology for IoT networks. In wireless-powered IoT networks, broadcasting is a fundamental networking service for disseminating messages to the whole network. To seek a fast and collision-free broadcast schedule, the problem of Minimum Latency Broadcast Scheduling (MLBS) has been well studied when nodes are energy-abundant. However, in wireless-powered networks, a node can only receive or transmit packets after it has harvested enough energy. In such networks, it is of great importance to exploit the divergent harvested energy to reduce the broadcast latency. Unfortunately, existing works always assume a predetermined tree and a fixed transmission power for broadcast scheduling, which greatly limits their performance. Thus, in this article, we investigate the first work for the MLBS problem in wireless-powered networks without relying on predetermined trees. First, the problem is formulated and proved to be NP-hard. Then, two structure-adaptive scheduling algorithms are proposed with a theoretical bound, which can intertwine the construction of broadcast tree with the computation of an energy-aware schedule simultaneously. Furthermore, a power-aware scheduling method is also proposed to take the structure of the broadcast tree, the adjustment of nodes’ transmission powers, and the interference during transmissions into account simultaneously. Additionally, the algorithm for the MLBS problem under the physical interference model is also studied. Finally, the theoretical analysis and simulation results verify that the proposed algorithms have high performance in terms of latency.

Mobile-edge computing (MEC) has received wide attention recently due to its efficacy in alleviating the computation stress of mobile devices (MDs), which is realized by offloading workloads from MD users to nearby edge servers (ESs). Prior work has studied related task offloading and scheduling problems and proposed many approaches. However, none of these approaches considers the reliability issue in MEC systems which may suffer soft errors during task execution as well as bit errors during task offloading simultaneously. Targeting optimization on a multi-user MEC system, in this article we investigate the task offloading and scheduling problem of maximizing system quality of experience (QoE) under a certain reliability requirement. With the consideration of the combinatorial nature of this problem, we propose to decompose the original problem into i) a task-to-ES assignment problem with fixed task offloading decision, for satisfying system reliability constraint, ii) a computing resource allocation problem with fixed task offloading and assignment decisions, for maximizing system QoE, and iii) a task offloading optimization problem to find the best offloading decision that achieves the maximum QoE under the reliability constraint using our task assignment and resource allocation methods. In order to solve these sub-problems, we further design a reliability-aware task-to-ES assignment algorithm, a QoE-optimum resource allocation algorithm, and a binary particle swarm optimization based task offloading algorithm. We perform extensive simulations and testbed experiments to validate the efficacy of the proposed scheme. Simulation and testbed results show that the proposed scheme greatly outperforms four benchmark approaches and it achieves up to 63.2% and 43.1% increase in the average QoE (quantified by offloading utility), respectively.

Open-World Compositional Zero-shot Learning (OW-CZSL) aims to recognize novel compositions of state and object primitives in images with no priors on the compositional space, which induces a tremendously large output space containing all possible state-object compositions. Existing works either learn the joint compositional state-object embedding or predict simple primitives with separate classifiers. However, the former method heavily relies on external word embedding methods, and the latter ignores the interactions of interdependent primitives, respectively. In this paper, we revisit the primitive prediction approach and propose a novel method, termed Progressive Cross-primitive Compatibility (ProCC), to mimic the human learning process for OW-CZSL tasks. Specifically, the cross-primitive compatibility module explicitly learns to model the interactions of state and object features with the trainable memory units, which efficiently acquires cross-primitive visual attention to reason high-feasibility compositions, without the aid of external knowledge. Moreover, to alleviate the invalid cross-primitive interactions, especially for partial-supervision conditions (pCZSL), we design a progressive training paradigm to optimize the primitive classifiers conditioned on pre-trained features in an easy-to-hard manner. Extensive experiments on three widely used benchmark datasets demonstrate that our method outperforms other representative methods on both OW-CZSL and pCZSL settings by large margins.

The explosive growth of video traffic on today's Internet promotes the rise of Neural-enhanced Video Streaming (NeVS), which effectively improves the rate-distortion trade-off by employing a cheap neural super-resolution model for quality enhancement on the receiver side. Missing by existing work, we reveal that the NeVS pipeline may suffer from a practical threat, where the crucial codec component (i.e., encoder for compression and decoder for restoration) can trigger adversarial attacks in a man-in-the-middle manner to significantly destroy video recovery performance and finally incurs the malfunction of downstream video perception tasks. In this paper, we are the first attempt to inspect the vulnerability of NeVS and discover a novel adversarial attack, called codec hijacking, where the injected invisible perturbation conspires with the malicious encoding matrix by reorganizing the spatial-temporal bit allocation within the bitstream size budget. Such a zero-day vulnerability makes our attack hard to defend because there is no visual distortion on the recovered videos until the attack happens. More seriously, this attack can be extended to diverse enhancement models, thus exposing a wide range of video perception tasks under threat. Evaluation based on state-of-the-art video codec benchmark illustrates that our attack significantly degrades the recovery performance of NeVS over previous attack methods. The damaged video quality finally leads to obvious malfunction of downstream tasks with over 75% success rate. We hope to arouse public attention on codec hijacking and its defence.

Federated learning (FL) has emerged as a powerful paradigm for learning from decentralized data, and federated domain generalization further considers the test dataset (target domain) is absent from the decentralized training data (source domains). However, most existing FL methods assume that domain labels are provided during training, and their evaluation imposes explicit constraints on the number of domains, which must strictly match the number of clients. Because of the underutilization of numerous edge devices and additional cross-client domain annotations in the real world, such restrictions may be impractical and involve potential privacy leaks. In this paper, we propose an efficient and novel approach, called Disentangled Prompt Tuning (DiPrompT), a method that tackles the above restrictions by learning adaptive prompts for domain generalization in a distributed manner. Specifically, we first design two types of prompts, i.e., global prompt to capture general knowledge across all clients and domain prompts to capture domain-specific knowledge. They eliminate the restriction on the one-to-one mapping between source domains and local clients. Furthermore, a dynamic query metric is introduced to automatically search the suitable domain label for each sample, which includes two-substep text-image alignments based on prompt tuning without labor-intensive annotation. Extensive experiments on multiple datasets demonstrate that our DiPrompT achieves superior domain generalization performance over state-of-the-art FL methods when domain labels are not provided, and even outperforms many centralized learning methods using domain labels.

Existing Knowledge Distillation (KD) methods typically focus on transferring knowledge from a large-capacity teacher to a low-capacity student model, achieving sub-stantial success in unimodal knowledge transfer. However, existing methods can hardly be extended to Cross-Modal Knowledge Distillation (CMKD), where the knowledge is transferred from a teacher modality to a different student modality, with inference only on the distilled student modality. We empirically reveal that the modality gap, i.e., modality imbalance and soft label misalignment, incurs the in-effectiveness of traditional KD in CMKD. As a solution, we propose a novel f_ustomized Crossmodal Knowledge Distillation (C2KD). Specifically, to alleviate the modality gap, the pre-trained teacher performs bidirectional distillation with the student to provide customized knowledge. The On-the-Fly Selection Distillationi OFSD) strategy is applied to selectively filter out the samples with misaligned soft labels, where we distill cross-modal knowledge from non-target classes to avoid the modality imbalance issue. To further provide receptive cross-modal knowledge, proxy student and teacher, inheriting unimodal and cross-modal knowledge, is formulated to progressively transfer cross-modal knowledge through bidirectional distillation. Experimental results on audio-visual, image-text, and RGB-depth datasets demonstrate that our method can effectively transfer knowledge across modalities, achieving superior performance against traditional KD by a large margin.

The training of multi-agent reinforcement learning (MARL) tasks with the public goods dilemma (PGD) is difficult because the selfish actions of individual agents for high personal rewards may reduce the collective utility of the whole group. Existing solutions to this problem, e.g., reward gifting or intrinsic rewards, although inducing cooperation among agents in small groups, cannot guarantee fairness among agents’ policies and fail to achieve optimal group utility in large-scale systems. In this paper, we propose F4PGD, an effective method to train large-scale MARL tasks with PGD in a decentralized manner, which is inspired by Adam’s equity theory that the match between a person’s payoff and his contribution is the key incentive for people to contribute to the common good. In F4PGD, a mechanism is designed to match an agent’s reward with its contribution, which suppresses agents from taking a free ride and meanwhile encourages well-learned agents to contribute to public goods. Experimental results show that F4PGD effectively learns optimal policies for the whole group and guarantees fairness among agents in several typical MARL tasks with PGD.

Serverless ML inference is an emerging cloud computing paradigm for low-cost, easy-to-manage inference services. In serverless ML inference, each call is executed in a container; however, the cold start of containers results in long inference delays. Unfortunately, most existing works do not work well because they still need to load models into containers from scratch, which is the bottleneck based on our observations. Therefore, this paper proposes a low-latency serverless ML inference system called Optimus via a new container management scheme. Our key insight is that the loading of a new model can be significantly accelerated when using an existing model with a similar structure in a warm but idle container. We thus develop a novel idea of inter-function model transformation for serverless ML inference, which delves into models within containers at a finer granularity of operations, designs a set of in-container meta-operators for both CNN and transformer model transformation, and develops an efficient scheduling algorithm with linear complexity for a low-cost transformation strategy. Our evaluations on thousands of models show that Optimus reduces inference latency by 24.00% ~ 47.56% in both simulated and real-world workloads compared to state-of-the-art work.

Cloud-edge collaboration Architecture (CEA) enables the co-training of AI models by cloud servers and edge servers, offering a promising solution for large-scale model training. An efficient data trading mechanism helps encourage edges to invest data resources to participate in training while reducing the cost of cloud servers. Existing research on data trading within CEA focuses on static scenarios, either overlooking the dynamics of data demand and the fairness of the selected edges or assuming unknown future communication overheads. To bridge these gaps and consider the long-term fairness constraints, we propose an Online Data Trading mechanism for the CEA, called ODT, to improve the long-term utility. Technically, ODT decouples the long-term fairness constraint into a series of single time-slot sub-problems using the Lyapunov optimization method and applies dynamic programming to solve the single time-slot edge selection sub-problems. We prove the NP-hardness of the sub-problems, the performance bounds, and the computational complexity of the proposed algorithm. Evaluation results demonstrate that the proposed mechanism effectively improves long-term utility and achieves an efficient trade-off between fairness and utility.

Distributed Deep Neural Network (DNN) inference is a promising technology to explore the distributed resources in edge cloud to realize edge intelligence. Meanwhile the inherent resource sharing nature of edge cloud infrastructure also raises serious concerns on security and privacy. Software Guard Ex-tensions (SGX) emerges as a potential hardware-level solution but its limited secure memory (i.e., enclave page cache) imposes new challenges, especially in contrast to memory-hungry DNN models. A task's performance will be severely affected when its memory footprint is beyond the enclave page cache size, due to expensive secure page swapping. In this case, how to appropriately partition a DNN model and assign the partitions to distributed edge servers to efficiently utilize edge resources for fast secure inference becomes a challenging problem. In this paper, we first show that this problem is NP-hard. We further propose a MEmory -aware Distributed Inference Acceleration (MEDIA) algorithm, whose guaranteed approximation ratio is also formally analyzed. We have implemented a prototype system and applied some well-known representative DNN models to evaluate MEDIA's performance. Through extensive experiments, we verify the efficiency of MEDIA by the fact that it reduces the inference time by 19.5%-38.1 % in comparison with state-of-the-art approaches.

Personalized federated learning (PFL) has gained great success in tackling the scenarios where target datasets are heterogeneous across the local clients. However, the application of the existing PFL methods to real-world setting is hindered by the common assumption that the test data on each client is in-distribution (IND) with respect to its training data. Due to the bias of training dataset, the modern machine learning model prefers to rely on shortcut which can perform well on the training data but fail to generalize to the unseen test data that is out-of-distribution (OOD). This pervasive phenomenon is called shortcut learning and has attracted plentiful efforts in centralized situations. In PFL, the limited data diversity on federated clients makes mitigating shortcut and meanwhile preserving personalization knowledge rather difficult. In this paper, we analyse this challenging problem by formulating the structural causal models (SCMs) for heterogeneous federated clients. From the proposed SCMs, we derive two significant causal signatures which inspire a provable shortcut discovery and removal method under federated learning, namely FedSDR. Specifically, FedSDR is divided into two steps: 1) utilizing the available training data distributed among local clients to discover all the shortcut features in a collaborative manner. 2) developing the optimal personalized causally invariant predictor for each client by eliminating the discovered shortcut features. We provide theoretical analysis to prove that our method can draw complete shortcut features and produce the optimal personalized invariant predictor that can generalize to unseen OOD data on each client. The experimental results on diverse datasets validate the superiority of FedSDR over the state-of-the-art PFL methods on OOD generalization performance.

Fine-tuning the learnable prompt for a pre-trained vision-language model (VLM), such as CLIP, has demonstrated exceptional efficiency in adapting to a broad range of downstream tasks. Existing prompt tuning methods for VLMs do not distinguish spurious features introduced by biased training data from invariant features, and employ a uniform alignment process when adapting to unseen target domains. This can impair the cross-modal feature alignment when the testing data significantly deviate from the distribution of the training data, resulting in a poor out-of-distribution (OOD) generalization performance. In this paper, we reveal that the prompt tuning failure in such OOD scenarios can be attribute to the undesired alignment between the textual and the spurious feature. As a solution, we propose **CoOPood**, a fine-grained prompt tuning method that can discern the causal features and deliberately align the text modality with the invariant feature. Specifically, we design two independent contrastive phases using two lightweight projection layers during the alignment, each with different objectives: 1) pulling the text embedding closer to invariant image embedding and 2) pushing text embedding away from spurious image embedding. We have illustrated that **CoOPood** can serve as a general framework for VLMs and can be seamlessly integrated with existing prompt tuning methods. Extensive experiments on various OOD datasets demonstrate the performance superiority over state-of-the-art methods.

Exploiting invariant relations and mitigating spurious correlation (a.k.a., shortcut) between representation and target across varied data distributions can tackle the challenging out-of-distribution (OOD) generalization problem. In personalized federated learning (PFL), heterogeneous data distribution across local clients offers the inherent prerequisites to extract the invariant features that maintain invariant relation with target. Nevertheless, personalized features are closely entangled with spurious features in PFL since they exhibit similar variability across different clients, which makes preserving personalization knowledge and eliminating shortcuts two conflicting objectives in PFL. To address the above challenge, we analyse the heterogeneous data generation on local clients through the lens of structured causal model and propose a crucial causal signature which can distinguish personalized features from spurious features with global invariant features as the anchor. Then the causal signature is quantified as an information-theoretic constraint that facilitates the shortcut-averse personalized invariant learning on each client. Theoretical analysis demonstrates our method, FedPIN, can yield a tighter bound on generalization error than the prevalent PFL approaches when train-test distribution shift exists on clients. Moreover, we provide a theoretical guarantee on the convergence rate of FedPIN in this paper. The results of extensive experiments show that our method can achieve superior OOD generalization performance compared with the state-of-the-art competitors.

Recent diffusion models have manifested extraordinary capabilities in generating high-quality, diverse, and innovative images guided by textual prompts. Nevertheless, these state-of-the-art models may encounter the challenge of concept bleeding when generating images with multiple entities or attributes in the prompt, leading to the unanticipated merging or overlapping of distinct objects in the synthesized result. The current work exploits auxiliary networks to produce mask-constrained regions for entities, necessitating the training of an object detection network. In this paper, we investigate the bleeding reason and find that the cross-attention map associated with a specific entity or attribute tends to extend beyond its intended focus, encompassing the background or other unrelated objects and thereby acting as the primary source of concept bleeding. Motivated by this, we propose Entity Localization and Anchoring (ELA) to drive the entity to concentrate on the expected region accurately during inference, eliminating the necessity for training. Specifically, we initially identify the region corresponding to each entity and subsequently employ a tailored loss function to anchor entities within their designated positioning areas. Extensive experiments demonstrate its superior capability in precisely generating multiple objects as specified in the textual prompts.

Transformer model empowered architectures have become a pillar of cloud services that keeps reshaping our society. However, the dynamic query loads and heterogeneous user requirements severely challenge current transformer serving systems, which rely on pre-training multiple variants of a foundation model, i.e., with different sizes, to accommodate varying service demands. Unfortunately, such a mechanism is unsuitable for large transformer models due to the additional training costs and excessive I/O delay. In this paper, we introduce OTAS, the first elastic serving system specially tailored for transformer models by exploring lightweight token management. We develop a novel idea called token adaptation that adds prompting tokens to improve accuracy and removes redundant tokens to accelerate inference. To cope with fluctuating query loads and diverse user requests, we enhance OTAS with application-aware selective batching and online token adaptation. OTAS first batches incoming queries with similar service-level objectives to improve the ingress throughput. Then, to strike a tradeoff between the overhead of token increment and the potentials for accuracy improvement, OTAS adaptively adjusts the token execution strategy by solving an optimization problem. We implement and evaluate a prototype of OTAS with multiple datasets, which show that OTAS improves the system utility by at least 18.2%.

Recently, federated learning (FL) has gained momentum because of its capability in preserving data privacy. To conduct model training by FL, multiple clients exchange model updates with a parameter server via Internet. To accelerate the communication speed, it has been explored to deploy a programmable switch (PS) in lieu of the parameter server to coordinate clients. The challenge to deploy the PS in FL lies in its scarce memory space, prohibiting running memory consuming aggregation algorithms on the PS. To overcome this challenge, we propose Federated Learning in-network Aggregation with Compression (FediAC) algorithm, consisting of two phases: client voting and model aggregating. In the former phase, clients report their significant model update indices to the PS to estimate global significant model updates. In the latter phase, clients upload global significant model updates to the PS for aggregation. FediAC consumes much less memory space and communication traffic than existing works because the first phase can guarantee consensus compression across clients. The PS easily aligns model update indices to swiftly complete aggregation in the second phase. Finally, we conduct extensive experiments by using public datasets to demonstrate that FediAC remarkably surpasses the state-of-the-art baselines in terms of model accuracy and communication traffic.

Surveillance cameras are ubiquitous nowadays and users’ increasing needs for accessing real-world information (e.g., finding abandoned luggage) have urged object queries in real-time videos. While recent real-time video query processing systems exhibit excellent performance, they lack utility in deployment in practice as they overlook some crucial aspects, including multi-camera exploration, resource contention, and content awareness. Motivated by these issues, we propose a framework Gecko, to provide resource-efficient and accurate real-time object queries of massive videos on edge devices. Gecko (i) obtains optimal models from the model zoo and assigns them to edge devices for executing current queries, (ii) optimizes resource usage of the edge cluster at runtime by dynamically adjusting the frame query interval of each video stream and forking/joining running models on edge devices, and (iii) improves accuracy in changing video scenes by fine-grained stream transfer and continuous learning of models. Our evaluation with real-world video streams and queries shows that Gecko achieves up to 2x more resource efficiency gains and increases overall query accuracy by at least 12% compared with prior work, further delivering excellent scalability for practical deployment.

In Ethereum, the ledger exchanges messages along an underlying Peer-to-Peer (P2P) network to reach consistency. Understanding the underlying network topology of Ethereum is crucial for network optimization, security and scalability. However, the accurate discovery of Ethereum network topology is non-trivial due to its deliberately designed security mechanism. Consequently, existing measuring schemes cannot accurately infer the Ethereum network topology with a low cost. To address this challenge, we propose the Distributed Ethereum Network Analyzer (DEthna) tool, which can accurately and efficiently measure the Ethereum network topology. In DEthna, a novel parallel measurement model is proposed that can generate marked transactions to infer link connections based on the transaction replacement and propagation mechanism in Ethereum. Moreover, a workload offloading scheme is designed so that DEthna can be deployed on multiple distributed probing nodes so as to measure a large-scale Ethereum network at a low cost. We run DEthna on Goerli (the most popular Ethereum test network) to evaluate its capability in discovering network topology. The experimental results demonstrate that DEthna significantly outperforms the state-of-the-art baselines. Based on DEthna, we further analyze characteristics of the Ethereum network revealing that there exist more than 50% low-degree Ethereum nodes that weaken the network robustness.

Federated Unlearning (FU) aims to delete specific training data from an ML model trained using Federated Learning (FL). However, existing FU methods suffer from inefficiencies due to the high costs associated with gradient recomputation and storage. This paper presents QuickDrop, an original and efficient FU approach designed to overcome these limitations. During model training, each client uses QuickDrop to generate a compact synthetic dataset, serving as a compressed representation of the gradient information utilized during training. This synthetic dataset facilitates fast gradient approximation, allowing rapid downstream unlearning at minimal storage cost. To unlearn some knowledge from the trained model, QuickDrop clients execute stochastic gradient ascent with samples from the synthetic datasets instead of the training dataset. The tiny volume of synthetic data significantly reduces computational overhead compared to conventional FU methods. Evaluations with three standard datasets and five baselines show that, with comparable accuracy guarantees, QuickDrop reduces the unlearning duration by 463× compared to retraining the model from scratch and 65 -- 218× compared to FU baselines. QuickDrop supports both class- and client-level unlearning, multiple unlearning requests, and relearning of previously erased data.

Recently, Multimodal Learning (MML) has gained significant interest as it compensates for single-modality limitations through comprehensive complementary information within multimodal data. However, traditional MML methods generally use the joint learning framework with a uniform learning objective that can lead to the modality competition issue, where feedback predominantly comes from certain modalities, limiting the full potential of others. In response to this challenge, this paper introduces DI-MML, a novel detached MML framework designed to learn complementary information across modalities under the premise of avoiding modality competition. Specifically, DI-MML addresses competition by separately training each modality encoder with isolated learning objectives. It further encourages cross-modal interaction via a shared classifier that defines a common feature space and employing a dimension-decoupled unidirectional contrastive (DUC) loss to facilitate modality-level knowledge transfer. Additionally, to account for varying reliability in sample pairs, we devise a certainty-aware logit weighting strategy to effectively leverage complementary information at the instance level during inference. Extensive experiments conducted on audio-visual, flow-image, and front-rear view datasets show the superior performance of our proposed method. The code is released at https://github.com/fanyunfeng-bit/DI-MML.

This paper provides an efficient training-free painterly image harmonization (PIH) method, dubbed FreePIH, that leverages only a pre-trained diffusion model to achieve state-of-the-art harmonization results. Unlike existing methods that require either training auxiliary networks or fine-tuning a large pre-trained backbone, or both, to harmonize a foreground object with a painterly-style background image, our FreePIH tames the denoising process as a plug-in module for foreground image style transfer. Specifically, we find that the very last few steps of the denoising (i.e., generation) process strongly correspond to the stylistic information of images, and based on this, we propose to augment the latent features of both the foreground and background images with Gaussians for a direct denoising-based harmonization. To guarantee the fidelity of the harmonized image, we make use of latent features to enforce the consistency of the content and stability of the foreground objects in the latent space, and meanwhile, aligning both fore-/back-grounds with the same style. Moreover, to accommodate the generation with more structural and textural details, we further integrate text prompts to attend to the latent features, hence improving the generation quality. Quantitative and qualitative evaluations on COCO and LAION 5B datasets demonstrate that our method can surpass representative baselines by large margins.

Recent studies reveal that even highly biased dense networks can contain an invariant substructure with superior out-of-distribution (OOD) generalization. While existing works commonly seek these substructures using global sparsity constraints, the uniform imposition of sparse penalties across samples with diverse levels of spurious contents renders such methods suboptimal. The precise adaptation of model sparsity, specifically tailored for spurious features, remains a significant challenge. Motivated by the insight that in-distribution (ID) data containing spurious features may exhibit lower experiential risk, we propose a novel Spurious Feature-targeted Pruning framework, dubbed SFP, to induce the authentic invariant substructures without referring to the above concerns. Specifically, SFP distinguishes spurious features within ID instances during training by a theoretically validated threshold. It then penalizes the corresponding feature projections onto the model space, steering the optimization towards subspaces spanned by those invariant factors. Moreover, we also conduct detailed theoretical analysis to provide a rationality guarantee and a proof framework for OOD structures based on model sparsity. Experiments on various OOD datasets show that SFP can significantly outperform both structure-based and non-structure-based OOD generalization state-of-the-art (SOTA) methods by large margins.

Recent advancements in vision-language-to-image (VL2I) diffusion generation have made significant progress. While generating images from broad vision-language inputs holds promise, it also raises concerns about potential misuse, such as copying artistic styles without permission, which could have legal and social consequences. Therefore, it's crucial to establish governance frameworks to ensure ethical and copyright integrity, especially with widely used diffusion models. To address these issues, researchers have explored various approaches, such as dataset filtering, adversarial perturbations, machine unlearning, and inference-time refusals. However, these methods often lack either scalability or effectiveness. In response, we propose a new framework called causal representation editing (CRE), which extends representation editing from large language models (LLMs) to diffusion-based models. CRE enhances the efficiency and flexibility of safe content generation by intervening at diffusion timesteps causally linked to unsafe concepts. This allows for precise removal of harmful content while preserving acceptable content quality, demonstrating superior effectiveness, precision and scalability compared to existing methods. CRE can handle complex scenarios, including incomplete or blurred representations of unsafe concepts, offering a promising solution to challenges in managing harmful content generation in diffusion-based models.

Multi-modal domain generalization (MMDG) requires that models trained on multi-modal source domains can generalize to unseen target distributions with the same modality set. Sharpness-aware minimization (SAM) is an effective technique for traditional uni-modal domain generalization (DG), however, with limited improvement in MMDG. In this paper, we identify that modality competition and discrepant uni-modal flatness are two main factors that restrict multi-modal generalization. To overcome these challenges, we propose to construct consistent flat loss regions and enhance knowledge exploitation for each modality via cross-modal knowledge transfer. Firstly, we turn to the optimization on representation-space loss landscapes instead of traditional parameter space, which allows us to build connections between modalities directly. Then, we introduce a novel method to flatten the high-loss region between minima from different modalities by interpolating mixed multi-modal representations. We implement this method by distilling and optimizing generalizable interpolated representations and assigning distinct weights for each modality considering their divergent generalization capabilities. Extensive experiments are performed on two benchmark datasets, EPIC-Kitchens and Human-Animal-Cartoon (HAC), with various modality combinations, demonstrating the effectiveness of our method under multi-source and single-source settings. Our code is open-sourced.

Federated Knowledge Graph Embedding (FKGE) is an emerging collaborative learning technique for deriving expressive representations (i.e., embeddings) from client-maintained distributed knowledge graphs (KGs). However, poisoning attacks in FKGE, which lead to biased decisions by downstream applications, remain unexplored. This paper is the first work to systematize the risks of FKGE poisoning attacks, from which we develop a novel framework for poisoning attacks that force the victim client to predict specific false facts. Unlike centralized KGEs, FKGE maintains KGs locally, making direct injection of poisoned data challenging. Instead, attackers must create poisoned data without access to the victim's KG and inject it indirectly through FKGE aggregation. Specifically, to create poisoned data, the attacker first infers the targeted relations in the victim's local KG via a new KG component inference attack. Then, to accurately mislead the victim's embeddings via aggregation, the attacker locally trains a shadow model using the poisoned data and uses an optimized dynamic poisoning scheme to adjust the model and generate progressive poisoned updates. Our experimental results demonstrate the attack's effectiveness, achieving a remarkable success rate on various KGE models (e.g., 100% on TransE with WN18RR) while keeping the original task's performance nearly unchanged.

In cloud-based health monitoring services, healthcare centers often outsource support vector machine (SVM)-based clinical decision models to provide remote users with clinical decisions. During service provisioning, authorized external organizations like insurance companies aim to verify decision correctness to prevent fraudulent medical reimbursements. However, existing verifiable and secure SVM classification schemes have predominantly focused on user self-verification, thereby introducing potential risks of privacy leakage (such as input data exposure) in publicly verifiable scenarios. To address the aforementioned limitation, we propose a publicly verifiable and secure SVM classification scheme (PVSSVM) for cloud-based health monitoring services in a malicious setting, which can accommodate the verification needs of users or authorized external organizations with respect to potential malicious results returned by cloud servers. Specifically, we utilize homomorphic encryption and secret sharing to protect the model and data confidentiality in the cloud server, respectively. Based on a multiserver verifiable computation framework, PVSSVM achieves public verification of predicted results. Additionally, we further investigate its performance. Experimental evaluations demonstrate that PVSSVM outperforms existing state-of-the-art solutions in terms of computation and communication overhead. Notably, in the verification scenario of large-scale predictions, the proposed scheme achieves a reduction of approximately 83.71% in computation overhead through batch verification, as compared to one-by-one verification.

In hierarchical federated learning (HFL), edge computing is introduced for partial model aggregation to reduce latency. High energy cost is an important issue to be solved in the process of parameters uploading. Our study focuses on the issue of minimizing energy cost with delay constraint through node cooperation in HFL, and the decision problem for this is NP-hard. We introduce a cost-efficient HFL (CE-HFL) framework, where nodes not only participate in model training but also transmit and aggregate model parameters for neighbors. A parameter aggregation tree is first generated, and parameter updates can be delivered to edge servers along paths in the tree while being aggregated simultaneously. Through theoretical analysis, it is proved that CE-HFL can achieve energy optimization with delay constraint. We also evaluate its performance through thorough experiments. In comparison with HierFAVG, CFL, and HFEL, it is found that CE-HFL can save energy cost up to 24.58%, 22.02%, and 6.60%, respectively.

Federated learning (FL) as an emerging distributed machine learning (ML) paradigm enables participants to train their on-device data locally and share model parameters with others by the parameter server. Differing from the centralized ML, FL splits the high requirements of training data and computing power from the server to clients, which is well adapted to unmanned aerial vehicle (UAV) swarms with scattered nodes, heterogeneous data, and limited computing power. However, pre-trained models are unsatisfactory in unfamiliar scenes and most existing approaches fail to concentrate on the communication-sensitivity and real-time requirements in UAV-enabled FL scenarios. To address this problem, this paper proposes participant and sample selection for efficient online federated learning in UAV swarms (FedOL). Through the combination of online learning and FL, UAVs can supplement real-time samples and quickly improve the model accuracy in unfamiliar scenes. Meanwhile, to reduce the training latency with expected model accuracy, FedOL allows the server UAV to select participants with high training utility, while the client UAVs select more important samples. We implement FedOL and deploy it on UAV embedded devices. Experimental results show that compared with existing FL approaches, FedOL speeds up by about 2.61× and reaches the final accuracy about 1.02× higher.

In the Industrial Internet of Things (IIoT), cross-silo federated learning (CSFL) enables entities, such as manufacturers and suppliers to train global models for optimizing production processes while ensuring data privacy. A well-designed incentive mechanism is essential to persuade clients to contribute data resources. However, existing methodologies overlook the dynamic nature of the training process, where the accuracy of the globally trained model and the client’s data ownership change over time. Furthermore, the majority of previous research assumes a defined functional relationship between the data contribution and the model accuracy, which is infeasible in realistic and dynamic training environments. To address these challenges, we design a novel adaptive mechanism for CSFL that inspires organizations to contribute data resources in a dynamic training environment with the aim of maximizing their long-term payoffs. This mechanism leverages multiagent reinforcement learning (MARL) to ascertain near-optimal data contribution strategies from potential game histories without necessitating private organizational information or a precise accuracy function. Experimental results indicate that our mechanism achieves adaptive incentive in dynamic environments and effectively enhances the long-term payoffs of organizations.

Metaverse, a hypothetical digital environment linking the cyber world and the physical world, is expected to revolutionize the way people interact. In the metaverse, people interact with objects, the environment, and each other through digital representations of themselves or avatars across time and space. For example, in the metaverse, people can have meetings with colleagues hundreds of miles away. They can also walk through the aisles of a store, find the best fit and have it delivered to their doorstep. It is also possible to simulate the optimal process manufacturing line to adjust for product variation and minimize bottlenecks, or test an innovative aircraft wing design without building expensive prototypes.

Metaverse, a hypothetical digital environment linking the cyber world and the physical world, is expected to revolutionize the way people interact. In the metaverse, people interact with objects, the environment, and each other through digital representations of themselves or avatars across time and space. For example, in the metaverse, people can have meetings with colleagues hundreds of miles away. They can also walk through the aisles of a store, find the best fit, and have it delivered to their doorstep. It is also possible to simulate the optimal process manufacturing line to adjust for product variation and minimize bottlenecks, or test an innovative aircraft wing design without building expensive prototypes.

Federated learning (FL) has been extensively exploited in the training of machine learning models to preserve data privacy. In particular, wireless FL enables multiple clients to collaboratively train models by sharing model updates via wireless communication without exposing raw data. The state-of-the-art wireless FL advocates efficient aggregation of model updates from multiple clients by over-the-air computing. However, a significant deficiency of over-the-air aggregation lies in the infeasibility of TopK model compression given that top model updates cannot be aggregated directly before they are aligned according to their indices. In view of the fact that TopK can greatly accelerate FL, we design a novel wireless FL with voting based TopK algorithm, namely WFL-VTopK, so that top model updates can be aggregated by over-the-air computing directly. Specifically, there are two phases in WFL-VTopK. In Phase 1, clients vote their top model updates, based on which global top model updates can be efficiently identified. In Phase 2, clients formally upload global top model updates so that they can be directly aggregated by over-the-air computing. Furthermore, the convergence of WFL-VTopK is theoretically guaranteed under non-convex loss. Based on the convergence of WFL-VTopK, we optimize model utility subjecting to training time and energy constraints. To validate the superiority of WFL-VTopK, we extensively conduct experiments with real datasets under wireless communication. The experimental results demonstrate that WFL-VTopK can effectively aggregate models by only communicating 1%-2% top models updates, and hence significantly outperforms the state-of-the-art baselines. By significantly reducing the wireless communication traffic, our work paves the road to train large models in wireless FL.

For increasing Internet of Things (IoT) devices, 6G wireless technology aims for ubiquitous communications in which positioning services are necessary. Private spatial keyword-based positioning service is promising in 6G-oriented IoT since it positions users based on spatial locations and textual keywords while protecting user privacy. However, due to economic benefits or malicious attacks, positioning service providers may return erroneous or incomplete results, which cause tremendous economic damage and security threats, e.g., always assigning a selective driver for the specific car-hailing user. A technical challenge for extending existing private schemes to enable users to verify the correctness and completeness of positioning results is the distinctive positioning paradigm between compared spatial locations and matched textual keywords. This paper proposes a private and verifiable spatial keyword positioning scheme named VSpatial in 6G-oriented IoT. VSpatial enables users to verify the correctness and completeness of spatial keyword-based positioning results while preserving user privacy. The main inspiration for addressing the technical challenge is converting both spatial locations and textual keywords into an internal status, i.e., adapting comparison and matching to existence judging by multiple cryptographic tools, such as hierarchical cube and pseudorandom function. Based on this inspiration, we design a novel private authenticated data structure (named PVTree), and then propose two constructions of VSpatial, i.e., VSpatial-S and VSpatial-D, to suit static and dynamic environments, respectively. The core idea for adapting VSpatial-S to VSpatial-D is transferring one whole PVTree into multiple exponential-size partitions. Security analysis proves the security and verifiability of VSpatial. Theoretical and experimental evaluations show that VSpatial achieves faster-than-linear positioning efficiency and linear verification overhead.

Real-time video perception tasks are often challenging on resource-constrained edge devices due to the issues of accuracy drop and hardware overhead, where saving computations is the key to performance improvement. Existing methods either rely on domain-specific neural chips or priorly searched models, which require specialized optimization according to different task properties. These limitations motivate us to design a general and task-independent methodology, called Patch Automatic Skip Scheme (PASS), which supports diverse video perception settings by decoupling acceleration and tasks. The gist is to capture inter-frame correlations and skip redundant computations at patch level, where the patch is a non-overlapping square block in visual. PASS equips each convolution layer with a learnable gate to selectively determine which patches could be safely skipped without degrading model accuracy. Specifically, we are the first to construct a self-supervisory procedure for gate optimization, which learns to extract contrastive representations from frame sequences. The pre-trained gates can serve as plug-and-play modules to implement patch-skippable neural backbones, and automatically generate proper skip strategy to accelerate different video-based downstream tasks, e.g., outperforming state-of-the-art MobileHumanPose in 3D pose estimation and FairMOT in multiple object tracking, by up to 9.43 × and 12.19 × speedups, respectively, on NVIDIA Jetson Nano devices.

Decentralized identity (DID) systems conforming to the World Wide Web Consortium (W3C) Decentralized Identifiers (DIDs) and Verifiable Credentials Data Model recommendations have recently attracted attention due to their better autonomy, interoperability, and openness design. However, those W3C recommendations lack a design for addressing the single point of failure (SPOF) and identity revocation, which could seriously compromise the robustness and practicality of DID systems. To remedy these limitations, we propose FutureDID, a DID system that enables multiple parties to jointly issue credentials and efficiently revoke DID identities, providing a robust and practical DID system. FutureDID is designed with a multi-party credential issuing mechanism based on distributed key generation technology, which transforms trust from a single entity to distributed committees and facilitates authentication between issuers, making it more resistant to SPOF. Moreover, the underlying blockchain system is built on a chameleon hash function to ensure tamper-proof and enable efficient identity revocation. We have implemented a prototype system using FISCO BCOS and conducted extensive evaluations to demonstrate the effectiveness and practicality of our system. Our evaluations have shown that FutureDID provides a significant improvement in efficiency, achieving at least a 60 × efficiency improvement in identity revocation compared to state-of-the-art systems.

Driven by data and models, the digital twin technique presents a new concept of optimizing system design, process monitoring, decision-making and more, through performing comprehensive virtual-reality interaction and continuous mapping. By introducing serverless computing to Mobile Edge Computing (MEC) environments, the emerging serverless edge computing paradigm facilitates the communication-efficient digital twin services and promises agile, fine-grained and cost-efficient provisioning of limited edge resources, where serverless functions are implemented by containers in cloudlets (edge servers). However, the nonnegligible cold start delay of containers deteriorates the responsiveness of digital twin services dramatically and the perceived user service experience. In this paper, we investigate delay-sensitive query service provisioning in digital twin-empowered serverless edge computing by considering user mobility. With digital twins of users deployed in the remote cloud, referred to as primary digital twins, we deploy their digital twin replicas based on serverless functions in cloudlets to mitigate the query service delay while enhancing user service satisfaction that is expressed as a utility function. We study two optimization problems with the aim of maximizing the accumulative utility gain: the digital twin replica placement problem per time slot, and the dynamic digital twin replica placement problem over a finite time horizon. We first formulate an Integer Linear Program (ILP) solution for the digital twin replica placement problem when the problem size is small; otherwise, we propose an approximation algorithm for the problem with a provable approximation ratio. We then design an online algorithm for the dynamic digital twin replica placement problem, and a performance-guaranteed online algorithm for a special case of the problem by assuming each user issues a query at each time slot. Finally, we evaluate the performance of the proposed algorithms for placing digital twin replicas in MEC networks through simulations. The results demonstrate the proposed algorithms are promising, outperforming their counterparts.

As a key technology of 5G, network function virtualization enables each monolithic service to be divided into microservices, facilitating their deployment and management in edge environments. One of the most critical issues in 5G is how to support dynamically arriving mission-critical services with low-latency and high-reliability requirements in distributed edge environments. However, most existing works focus on how to provide reliable services without considering latency, and their heuristics struggle to cope with high-dimensional constraints and complex environments with heterogeneous infrastructure and services. In this paper, we propose a SafeDRL algorithm to resource-efficiently support these dynamically arriving services while meeting their reliability and latency requirements. Specifically, we first formulate the problem as an integer nonlinear programming and prove its NP-hardness. To tackle this problem, our SafeDRL algorithm captures delayed rewards in dynamic environments by reinforcement learning, and corrects constraint violations with high-quality feasible solutions based on expert intervention, and prunes unnecessary backup instances for optimality. The algorithm is proved to have a bounded approximation ratio in general cases. Extensive trace-driven simulations show that, compared with the state-of-the-art solution, SafeDRL can save resource costs by up to 49.32% and improve the service acceptance ratio by up to 55% with acceptable execution time.

Distributed machine learning (DML) has shown great promise in accelerating model training on multiple GPUs. To increase GPU utilization, a common practice is to let multiple learning jobs share GPU clusters, where the most fundamental and critical challenge is how to efficiently schedule these jobs on GPUs. However, existing works about DML job scheduling are constrained to settings with homogeneous GPUs. GPU heterogeneity is common in practice, but its influence on multiple DML job scheduling has been seldom studied. Moreover, DML jobs have internal structures that contain great parallelism potentials, which have not yet been fully exploited in the heterogeneous computing environment. In this paper, we proposeHare, a DML job scheduler that exploits both inter-job and intra-job parallelism in a heterogeneous GPU cluster.Hareadopts a relaxed fixed-scale synchronization scheme that allows independent tasks to be flexibly scheduled within a training round. Given full knowledge of job arrival time and sizes, we propose a fast heuristic algorithm to minimize the average job completion time and derive its theoretical bound is derived. Without prior knowledge of jobs, we propose an online algorithm based on the Heterogeneity-aware Least-Attained Service (HLAS) policy. We evaluateHareusing a small-scale testbed and a trace-driven simulator. The results show that it can outperform the state-of-the-art, achieving a performance improvement of about 2.94×.

Federated Learning (FL) enables geo-distributed clients to collaboratively train a learning model without exposing their private data. By only exposing local model parameters, FL well preserves data privacy of clients. Yet, it remains possible to recover raw samples from over frequently exposed parameters resulting in privacy leakage. Differentially private federated learning (DPFL) has recently been suggested to protect these parameters by introducing information noises. In this way, even if attackers get these parameters, they cannot exactly infer true parameters from these noisy information. Directly incorporating Differentially Private (DP) into FL however can severely affect model utility. In this paper, we present an optimized sparse response mechanism (OSRM) that seamlessly incorporates DP into FL to diminish privacy budget consumption and improve model accuracy. Through OSRM, each FL client only exposes a selected set of large gradients, so as not to waste privacy budgets in protecting valueless gradients. We theoretically derive the convergence rate of DPFL with OSRM under non-convex loss. Then, OSRM is optimized by minimizing the loss of the convergence rate. Based on analysis, we present an effective algorithm for optimizing OSRM. Extensive experiments are conducted with public datasets, including MNIST, Fashion-MNIST and CIFAR-10. The results suggest that OSRM can achieve the average improvement of accuracy by 18.42% as compared to state-of-the-art baselines with a fixed privacy budget.

In spatial crowdsourcing, location-based task recommendation schemes are widely used to match appropriate workers in desired geographic areas with relevant tasks from data requesters. To ensure data confidentiality, various privacy-preserving location-based task recommendation schemes have been proposed, as cloud servers behave semi-honestly. However, existing schemes reveal access patterns, and the dimension of the geographic query increases significantly when additional information beyond locations is used to filter appropriate workers. To address the above challenges, this paper proposes two efficient and privacy-preserving location-based task recommendation (EPTR) schemes that support high-dimensional queries and access pattern privacy protection. First, we propose a basic EPTR scheme (EPTR-I) that utilizes randomizable matrix multiplication and public position intersection test (PPIT) to achieve linear search complexity and full access pattern privacy protection. Then, we explore the trade-off between efficiency and security and develop a tree-based EPTR scheme (EPTR-II) to achieve sub-linear search complexity. Security analysis demonstrates that both schemes protect the confidentiality of worker locations, requester queries, and query results and achieve different security properties on access pattern assurance. Extensive performance evaluation shows that both EPTR schemes are efficient in terms of computational cost, with EPTR-II being$10^{3}\times$faster than the state-of-the-art scheme in task recommendation.

Recently, increasing personal data has been stored in blockchain databases, ensuring data integrity by consensus. Although transparent and immutable blockchains are mainly adopted, the need to deploy preferences on which users canreadandeditthe data is growing in importance. Based on chameleon hashes, recent blockchains support editability governance but can hardly prevent data breaches because the data is readable to all participants in plaintexts. This motivates us to propose NANO, the first permissioned blockchain database that provides downward compatible readability and editability governance (i.e., users who caneditthe data can alsoreadthe data). Two challenges are protecting policy privacy and efficiently revoking malicious users (e.g., users who abuse their editability privileges). The punchline is leveraging Newton's interpolation formula-based secret sharing to hide policies into polynomial parameters and govern the distribution of data decryption keys and chameleon hash trapdoors. Inspired by proxy re-encryption, NANO integrates unique user symbols into user keys, achieving linear user revocation overhead. Security analysis proves that NANO provides comprehensive privacy preservation under the chosen-ciphertext attack. Experiments on the FISCO blockchain platform demonstrate that compared with state-of-the-art related solutions, NANO achieves a 7× improvement on average regarding computational costs, gas consumption, and communication overhead.

Previous state-of-the-art studies have proposed various analytical models to understand the double-spending attacks (DSA) occurred in Proof-of-Work (PoW) based Blockchains. Although many insights behind the double-spending attacks have been disclosed, we still believe that advanced versions of DSA can be developed to create new threats for the PoW-based blockchains such as the Bitcoin blockchain. In this paper, we present two new types of double-spending attacks in the context of the PoW-based blockchain and discuss the insights behind them. By considering more practical network parameters, such as the number of confirmation blocks, the hashpower of the double-spending attacker, the amount of coins in the target transaction, and the network-status parameter, we first analyze the success probability of the conventional double-spending attack, named Naive DSA. Based on Naive DSA, we create two new adaptive DSA, i.e., the Adaptive DSA and the Reinforcement Adaptive DSA (RA-DSA). In our analytical models, a double-spending attack is converted into a Markov Decision Process. We then exploit the Stochastic Dynamic Programming (SDP) approach to obtain the optimal attack strategies under Adaptive DSA and RA-DSA. Numerical simulation results demonstrate the correlations between each critical network parameter and the expected attacker's reward. Through the proposed analytical models, we aim to alert the PoW-based blockchain ecosystem that the threat of double-spending attacks is still at a dangerous level. For example, our findings show that the attacker can launch a successful attack with a small hashpower proportion much lower than 51% under RA-DSA. IEEE

Purpose: This research aims to present a distributed localized manufacturing (DLM)-based personalized customization Supply Chain (PCSC) model with facility siting, with the objective of enhancing the sustainability level of PCSCs within the context of Industry 5.0. Design/methodology/approach: To accomplish the objectives, a DLM-based PCSC model is constructed, and the supply chain is optimized using a P-Median model with genetic algorithm. Furthermore, a hybrid simulation model, combining agent-based modeling and discrete event simulation, is employed to analyze and gather data on sustainable metrics. Findings: The implementation of the DLM-based PCSC model yields improvements in supply chain efficiency through cost reduction, risk mitigation, and enhanced responsiveness. By leveraging the decentralized manufacturing approach of DLM, organizations can minimize transportation distances, optimize resource utilization, and strengthen coordination among geographically dispersed facilities. Consequently, sustainability and overall performance within the supply chain are enhanced. Practical implications: This study offers valuable recommendations for stakeholders and managers, including the adoption of distributed local manufacturing, optimization of facility locations, and the integration of sustainability metrics analysis into decision-making processes. These measures contribute to the improvement of supply chain sustainability and performance. Originality/value: This article makes a significant contribution to the field by proposing a DLM-based PCSC model that incorporates facility siting. The employment of a hybrid simulation model presents an integrated approach to assessing supply chains. In addition, it expands the measurement of sustainability metrics and provides insights to enhance the sustainability and efficiency of PCSCs.

By combining edge computing and parallel computing, distributed edge computing has emerged as a new paradigm to exploit the booming IoT devices at the edge. To accelerate computation at the edge,i.e., the inference tasks for DNN-driven applications, the parallelism of both computation and communication needs to be considered for distributed edge computing, and thus, the problem of Minimum Latency joint Communication and Computation Scheduling (MLCCS) is proposed. However, existing works have rigid assumptions that the communication time of each device is fixed and the workload can be split arbitrarily small. Aiming at making the work more practical and general, the MLCCS problem without the above assumptions is studied in this paper. Firstly, the MLCCS problem under a general model is formulated and proved to be NP-hard. Secondly, a pyramid-based computing model is proposed to consider the parallelism of communication and computation jointly, which has an approximation ratio of$1+\delta$, where$\delta$is related to devices' communication rates. An interesting property under such a computing model is identified and proved,i.e., the optimal latency can be obtained under arbitrary scheduling order when all the devices share the same communication rate. When the workload cannot be split arbitrarily, an approximation algorithm with a ratio of at most$2\cdot (1+\delta )$is proposed. Additionally, for handling the dynamically changing network scenarios, several algorithms are also proposed accordingly. Finally, the theoretical analysis and simulation results verify that the proposed algorithm has high performance in terms of latency. Two testbed experiments are also conducted, which show that the proposed method outperforms the existing methods, reducing the latency by up to 29.2% for inference tasks at the edge.

The escalating proliferation of user generated contents such as videos and images are dominating the network traffic. The optimal strategy for mitigating backbone congestion and minimizing user request latency lies in prudent caching at edge stations within distributed autonomous networks, obviating the necessity to transmit data to the cloud. However, accurately caching content based on distributed autonomous networks requires elaborative collaboration between edge servers, which remains a great challenge, especially when content is highly dynamic and the storage resources of edge stations are limited. To tackle this challenge, this paper proposes a contextual bandit-based online caching algorithm for evaluating the optimal content hit rate reward, which can adapt to the constantly changing stream of emerging content. We build the content space, BS space, and a fine-grained space searching method to cache contents and corresponding edge stations. Furthermore, to perform collaborative caching and sharing between edges, we propose a federated autonomous multi-layer caching framework, whereby each server can locally learn the model for accurate caching and a synchronous mechanism is set up for global updating, further improving the hit rates. Finally, we perform theoretical proofs and simulations, demonstrating that our regret is sublinear and our caching algorithm outperforms several state-of-the-art algorithms.

Extracting features from image data on Internet of Things (IoT) devices to reduce the amount of data that needs to be uploaded to cloud/edge servers has received increasing attention. However, most of the existing related approaches suffer from two major limitations, (i) low performance and high network traffic, and (ii) a lot of storage resource consumption. To this end, we propose a resource-efficient feature extraction framework for image processing in IoT devices. The proposed framework consists of the edge-assisted extractor generation method and the NestE method. The extractor generated by the edge-assisted extractor generation method can extract the features required by the application, which can not only avoid the IoT device uploading useless feature data but also improve application performance. The proposed NestE generates a nonredundant subextractor by splitting the extractor into multiple subextractors, removing redundant subextractors, and nesting small-capacity subextractors in large-capacity subextractors in a parameter-sharing manner. Compared with deploying multiple independent subextractors on IoT devices, deploying the nonredundant multifunctional extractor can save considerable storage resources and switching overhead. Extensive experimental results show that the proposed framework reduces the storage footprint by approximately 90.7% and switching overhead by approximately 92.4% compared with deploying independent subextractors when using the classical principal component analysis algorithm.

Using pre-trained vision-language models like CLIP with federated training prompts has shown great potential in federated learning (FL) by offering significant benefits in computation, communication, and privacy over existing frameworks. However, existing researches overlook the internal mechanisms underlying federated prompt tuning and comply with the traditional context-unaware tuning mechanism. Our experiments, on the other hand, demonstrate that federated prompting is a data-efficient but data-sensitive paradigm, and therefore, the samples that involved in the prompt tuning process holds significant importance. To address the above issue, we proposeContext-awareFederatedPromptTuning (CaFPT), which facilitates the retrieval process by conditioning on the examples capable of activating the most pertinent knowledge inside the pre-trained models with information theory. Moving in this direction steers the behavior of pre-trained neurons precisely and improves performance on the local task. Informative vectors are built by pruning clients’ training data based on their$\mathcal {V}$-usableinformation. The study shows that these vectors can be updated and combined through operations like FedAVG, and the resulting model's behavior is steered accordingly on multiple clients’ tasks. Extensive experiments have demonstrated that informative vectors offer promising robustness, making it a simple yet effective way to enhance the performance of federated prompting.

Quick global aggregation of effective distributed parameters is crucial to federated learning (FL), which requires adequate bandwidth for parameters communication and sufficient user data for local training. Otherwise, FL may cost excessive training time for convergence and produce inaccurate models. In this paper, we propose a brand-new FL framework, PromptFL, that replaces the federated model training with the federated prompt training, i.e., let federated participants train prompts instead of a shared model, to simultaneously achieve the efficient global aggregation and local training on insufficient data by exploiting the power of foundation models (FM) in a distributed way. PromptFL ships an off-the-shelf FM, i.e., CLIP, to distributed clients who would cooperatively train shared soft prompts based on very few local data. Since PromptFL only needs to update the prompts instead of the whole model, both the local training and the global aggregation can be significantly accelerated. And FM trained over large scale data can provide strong adaptation capability to distributed users tasks with the trained soft prompts. We empirically analyze the PromptFL via extensive experiments, and show its superiority in terms of system feasibility, user privacy, and performance.

Iteration based collaborative learning (CL) paradigms, such as federated learning (FL) and split learning (SL), faces challenges in training neural models over the rapidly growing yet resource-constrained edge devices. Such devices have difficulty in accommodating a full-size large model for FL or affording an excessive waiting time for the mandatory synchronization step in SL. To deal with such challenge, we propose a novel CL framework which adopts an tree-aggregation structure with an adaptive partition and ensemble strategy to achieve optimal synchronization and fast convergence at scale. To find the optimal split point for heterogeneous clients, we also design a novel partitioning algorithm by minimizing the idleness during communication and achieving the optimal synchronization between clients. In addition, a parallelism paradigm is proposed to unleash the potential of optimum synchronization between the clients and server to boost the distributed training process without losing model accuracy for edge devices. Furthermore, we theoretically prove that our framework can achieve better convergence rate than state-of-the-art CL paradigms. We conduct extensive experiments and show that our framework is 4.6× in training speed as compared with the traditional methods, without compromising training accuracy.

Mobile Edge Computing (MEC) has been identified as a desirable computing paradigm that provides efficient and effective services for various applications, while meeting stringent service delay requirements. Orthogonal to the MEC computing paradigm, Network Function Virtualization (NFV) technology is another enabling technology that provides the network resource management with great flexibility and scalability, where the instances of Virtual Network Functions (VNFs) are deployed in edge servers as Service Function Chains (SFCs) for SFC-enabled services. Although reliable service provisioning in MEC environments is fundamentally important, the deployed VNF instances usually are not reliable, which can be affected by their software implementation, their execution duration, the workload among edge servers, and so on. Empowered by digital twin techniques, the states of VNF instances can be maintained by their digital twins in a real-time manner and their reliability can be accurately predicted through their digital twins. In this paper, we study digital twin-assisted, SFC-enabled reliable service provisioning in MEC networks by exploiting the dynamics of VNF instance reliability. We concentrate on two novel optimization problems of reliable service provisioning: the service cost minimization problem, and the dynamic service admission maximization problem. We first show their NP-hardness. We then formulate an Integer Linear Program (ILP) solution, and devise an approximation algorithm with a constant approximation ratio for the service cost minimization problem. We thirdly provide an ILP solution to the offline version of the dynamic service admission maximization problem. Built upon this offline ILP solution, we also develop an online algorithm with a provable competitive ratio for the problem, by adopting the primal-dual dynamic updating technique. We finally evaluate the performance of the proposed algorithms via simulations. Simulation results demonstrate that the proposed algorithms outperform their comparison benchmarks, and improve the performance of their comparison counterparts by no less than$10.2 \%$.

Digital twins are poised to enter our lives with Industry 4.0. The Digital Twin Network (DTN) paradigm is projected to deliver upon the promise of efficient collaboration among digital twins to enable complicated and systematic services across many domains, through depicting an overall picture of a group of physical objects. To achieve timely data processing of digital twins, Mobile Edge Computing (MEC) shifts the computational power towards the network edge, and network slicing is well-suited to bundle heterogeneous physical resources to build logical networks based on edge servers for accommodating DTNs. In light of this, in this paper we investigate DTN slicing-enabled service provisioning in MEC, where each DTN slice consists of one master digital twin and a set of worker digital twins, and each worker digital twin is synchronized through collecting data from a respective object periodically. The master digital twin aggregates the processed data from worker digital twins to model the DTN continuously for user query services, whilst meeting delay requirements of users. We capture the utility gain of a DTN slicing request based on the DTN model quality at its master digital twin that is impacted by the Age of Information (AoI), and we focus on two novel optimization problems: the utility maximization problem for a single DTN slicing request, and the dynamic utility maximization problem for multiple DTN slicing requests. We propose an approximation algorithm for the former, and an online algorithm with a provable competitive ratio for the latter. We also evaluate the performance of the proposed algorithms through simulations. Experimental results demonstrate that the proposed algorithms are promising, outperforming their counterparts by at least 10.2%.

Propelled by recent advances in Mobile Edge Computing (MEC) and the Internet of Things (IoT), the digital twin technique has been envisioned as a de-facto driving force to bridge the virtual and physical worlds through creating digital portrayals of physical objects. In virtue of the flourishing of edge intelligence and abundant IoT data, data-driven modelling facilitates the implementation and maintenance of digital twins, where simulations of physical objects are usually performed based on Deep Neural Networks (DNNs). A significant advantage of adopting digital twins is to enable decisive prediction on the behaviours of objects in near future without waiting for that really happen. To provide accurate predictions, it is vital to keep each digital twin synchronized with its physical object in real-time. However, it is challenging to maintain the real-time synchronization between a digital twin and its physical object due to the dynamics of physical objects and sensing data drift over time, i.e., the live data from a physical object diverge from the model training data of its digital twin. To address this critical issue, continual learning is a promising solution to retrain models of digital twins incrementally. In this paper, we investigate digital twin synchronization issues via continual learning in an MEC environment, with the aim to maximize the total utility gain, i.e., the enhanced model accuracy. We study two novel optimization problems: the static digital twin synchronization problem per time slot and the dynamic digital twin synchronization problem for a finite time horizon. We first formulate an Integer Linear Program (ILP) solution for the static digital twin synchronization problem when the problem size is small; otherwise, we develop a randomized approximation algorithm at the expense of bounded resource violations for it. We also devise a deterministic approximation algorithm with guaranteed performance for a special case of the static digital twin synchronization problem. We thirdly consider the dynamic digital twin synchronization problem by proposing an efficient online algorithm for it. Finally, we evaluate the performance of the proposed algorithms for continuous digital twin synchronization through simulations. Simulation results show that the proposed algorithms are promising, outperforming counterpart benchmarks by no less than 13.2%, in terms of the total utility gain.

Over the past few years, edge learning has achieved significant success in mobile edge networks. Few works have designed incentive mechanism that motivates edge nodes to participate in edge learning. However, most existing works only consider myopic optimization and assume that all edge nodes are honest, which lacks long-term sustainability and the final performance assurance. In this paper, we propose Chiron, an incentive-driven Byzantine-resistant long-term mechanism based on hierarchical reinforcement learning (HRL). First, our optimization goal includes both learning-algorithm performance criteria (i.e., global accuracy) and systematical criteria (i.e., resource consumption), which aim to improve the edge learning performance under a given resource budget. Second, we propose a three-layer HRL architecture to handle long-term optimization, short-term optimization, and byzantine resistance, respectively. Finally, we conduct experiments on various edge learning tasks to demonstrate the superiority of the proposed approach. Specifically, our system can successfully exclude malicious nodes and lazy nodes out of the edge learning participation and achieves 14.96% higher accuracy and 12.66% higher total utility than the state-of-the-art methods under the same budget limit.

Performing accurate sensing in diverse environments is a challenging issue in wireless sensing technologies. Existing solutions usually require collecting a large number of samples to train a classifier for every environment, or further assume similar sample distribution between different environments such that a model trained in one environment can be transferred to another. In this paper, we propose RF-Siamese, an RFID-based gesture sensing approach that achieves comparable accuracy to existing solutions but requires only a few samples in each eivironment. RF-Siamese leverages Siamese networks to distinguish different gestures with only a small number of samples and is enhanced by several novel designs to achieve high accuracy in diverse environments. First, the network structure and parameters (e.g., loss function and distance metric) are carefully designed to be suitable for RFID gesture recognition. Second, a permutation-based dataset generation strategy is proposed to make full use of the collected samples to enhance the recognition accuracy. Third, a template matching method is proposed to extend the Siamese network to classify multiple gestures. Extensive experiments on commercial RFID devices demonstrate that RF-Siamese achieves a high accuracy of 0.93 with only one sample of each gesture when recognizing 18 different gestures, while state-of-the-art approaches based on transfer learning and meta learning achieve an accuracy of only 0.59 and 0.70, respectively.

Federated learning (FL) can train collaboratively with several mobile terminals (MTs), which faces critical challenges in communication, resource, and privacy. Existing privacy-preserving methods usually adopt instance-level differential privacy (DP), which provides a rigorous privacy guarantee but with several bottlenecks: performance degradation, transmission overhead, and resource constraints. Therefore, we propose Fed-LTP, an efficient and privacy-enhanced FL framework withLotteryTicketHypothesis (LTH) and zero-concentrated DP(zCDP). It generates a pruned global model on the server side and conducts sparse-to-sparse training from scratch with zCDP on the client side. On the server side, two pruning schemes are proposed: (i) the weight-based pruning (LTH) determines the pruned global model structure; (ii) the iterative pruning further shrinks the size of the pruned model. Meanwhile, the performance of Fed-LTP is boosted via model validation based on the Laplace mechanism. On the client side, we use sparse-to-sparse training to solve the resource-constraints issue and provide tighter privacy analysis to reduce the privacy budget. We evaluate the effectiveness of Fed-LTP on several real-world datasets in both independent and identically distributed (IID) and non-IID settings. The results confirm the superiority of Fed-LTP over state-of-the-art (SOTA) methods in communication, computation, and memory efficiencies while realizing a better utility-privacy trade-off.

Unmanned aerial vehicles (UAVs) play an essential role in emergency cases and adverse environments for applications like disaster detection and mine exploration. To process the massive volume of sensing data generated by various sensory payloads in these applications, existing works either compress deep learning (DL) models to conduct onboard computing, or offload raw data back to the resourceful ground station with the help of relay UAVs due to base station damage. However, the former sacrifices the inference accuracy of DL models (up to 10% accuracy loss), while the latter achieves high accuracy at the cost of significant latency, due to limited wireless communication resources in the multi-hop transmission. To address the problem, exploiting the resources of the UAV swarm including both task UAVs and relay UAVs, we build up anall-skyautonomous computing (ASAP) system to autonomously conduct collaborative computing in the swarm, to achieve both high accuracy and low latency of sensing data processing. In detail, we first propose a novel UAV swarm-native collaborative computing architecture, considering the general hierarchy and clustering structure of UAV swarms, as well as the characteristic of DL model execution. We then design an elastic efficient task scheduler to allocate computing tasks for UAVs, and update the scheduling scheme online when some UAVs are unavailable, with the aid of a lightweight and accurate DL inference performance predictor. Finally, we design an adaptive inter-UAV data compressor, to adapt to the limited and dynamic communication resources between UAVs. Experiment results on 24 airborne computers and five real-world UAVs show that, the proposed system can perform collaborative computing in a timely manner and effectively deal with situations when some UAVs become unavailable.

Collaborative tasks often require the cooperation of multiple individuals to be completed in mobile crowdsensing (MCS). However, previous task recommendations predominantly focused on individuals rather than groups, making them less effective for collaborative tasks. It is crucial to study the collaborative task recommendation problem in MCS. In this work, we propose an Attention-based Neural Collaborative approach (ANC) for group task recommendation. In particular, a grouping method is designed based on participant abilities to form groups that meet the needs of collaborative tasks. Meanwhile, a dual-attention mechanism is constructed to aggregate member preferences and enhance the representation of tasks and groups. The neural network-based collaborative filter mechanism is employed to generate top-$K$recommendation lists. Experimental results, based on two real-world datasets, demonstrate that ANC outperforms others, validating its effectiveness and feasibility.

Federated Learning (FL) has gained considerable attention recently, as it allows clients to cooperatively train a global machine learning model without sharing raw data. However, its performance can be compromised due to the high heterogeneity in clients' local data distributions, commonly known as Non-IID (non-independent and identically distributed). Moreover, collaboration among highly dissimilar clients exacerbates this performance degradation. Personalized FL seeks to mitigate this by enabling clients to collaborate primarily with others who have similar data characteristics, thereby producing personalized models. We noticed that existing methods for assessing model similarity often do not capture the genuine relevance of client domains. In response, our paper enhances personalized client collaboration in FL by introducing a metric for domain relevance between clients. Specifically, to facilitate optimal coalition formation, we measure the marginal contributions of client models using coalition game theory, providing a more accurate representation of potential client domain relevance within the FL privacy-preserving framework. Based on this metric, we then adjust each client's coalition membership and implement a personalized FL aggregation algorithm that is robust to Non-IID data domain. We provide a theoretical analysis of the algorithm's convergence and generalization capabilities. Our extensive evaluations on multiple datasets, including MNIST, Fashion-MNIST, CIFAR-10, and CIFAR-100, and under varying Non-IID data distributions (Pathological and Dirichlet), demonstrate that our personalized collaboration approach consistently outperforms contemporary benchmarks in terms of accuracy for individual clients.

Federated Learning (FL) system needs to incentivize clients since they may be reluctant to participate in the resource consuming process. Existing incentive mechanisms fail to construct a sustainable environment for the long-term development of FL system: 1) They seldom focus on system economic properties (e.g., social welfare, individual rationality, and incentive compatibility) to guarantee client attraction. 2) Current online auction modeling methods divide the whole continual process into multiple independent rounds and solve them one-by-one, which breaks the correlation between each round. Besides, the inherent characteristics of FL system (model-agnostic and privacy-sensitive) also prevent it from the optimal strategy by precise mathematical analysis. 3) Current system modelings ignore the practical problem of periodical client shifting, which cannot adaptively update its strategy to handle system dynamics. To overcome the above challenges, this paper proposes a long-term adaptive Vickrey–Clarke–Groves (VCG) auction mechanism for FL system, which incorporate a multi-branch deep reinforcement learning (DRL) algorithm. First, VCG auction is the only one that can simultaneously guarantee all crucial economic properties. Second, we extend the economic properties to long-term forms and apply the experience-driven DRL algorithm to directly obtain long-term optimal strategy, without any prior system knowledge. Third, we reconstruct a multi-branch DRL network to accommodate periodical client shifting by adaptive decision head switching for different time periods. Finally, we theoretically prove he extended economic properties (i.e., IC) and conduct extensive experiments on several real-world datasets. Compared with state-of-the-art approaches, the long-term social welfare of FL system increases by 36% with a 37% reduction in payment. Besides, the multi-branch network can adaptively handle periodical client shifting on the timeline.

WiFi-based wireless sensing has gained popularity for enabling smart indoor services, one of which is fall detection which plays a vital role in mitigating health risks for elders. Previous approaches have treated daily activities as independent events and built models to distinguish falls from others. However, human activities are usually adjoint in practice, e.g., the elder may suddenly fall when she walks. This adjoining introduces shared features between different activities, thereby affecting the classification performance. To address this problem, we propose Fall-attention, an attention-based fall detection method that can focus on the features related to fall events and suppress interference of irrelevant activities to improve performance. Its basic idea is to produce a task-oriented feature representation of fall events inside the signal using attention-based sentence embedding techniques and Recurrent Neural Network (RNN). We incorporate multi-task learning into Fall-attention by adopting multiple independent classification modules. This enables the model to explore different regions of the signal, capturing the composition of adjoint activities. A series of signal preprocessing and data enhancement techniques are also adopted to promote model training. Experimental results of the dataset containing adjoint activities demonstrate the superiority of Fall-attention over previous methods, which achieves an average accuracy of 95%.

Machine learning models are widely applied for vehicular applications, which are essential to future intelligent transportation system (ITS). Traditional model training methods commonly employ a client-server architecture to perform local training and global iterative aggregations, which can consume significant bandwidth resources that are often absent in vehicular networks, especially in high vehicle density scenarios. Modern vehicle users naturally can collaboratively train machine learning models as they are the data owner and have strong local computing power from the onboard units (OBU). In this paper, we propose a novel collaborative learning scheme for mobile vehicles that can utilize the opportunistic vehicle-to-roadside (V2R) communication to exploit the common priors of vehicular data without interaction with a centralized coordinator. Specifically, vehicles perform local training during the driving journey, and simply upload its local model to roadside unit (RSU) encountered on the way. RSU's model will be updated accordingly and sent back to the vehicle via the V2R communication. We have theoretically shown that RSUs' models can eventually converge without a backhaul connection. Extensive experiments upon various road configurations demonstrate that the proposed scheme can efficiently train models among vehicles without dedicated Internet access and scale well with both the road range and vehicle density.

Packet loss inferring can enable a transceiver to distinguish between channel impairment and collision for transmission failures, and thus can improve the network performance by exclusively performing rate adaptation or adjusting the medium access parameter. Machine learning methods from literature have shown great potential in producing models that can detect the loss causes over various network trace, however haven't considered accurate data-driven loss inferring on resource-constrained devices that cannot accommodate deep models. In this paper, we propose a novel packet loss inferring framework that can train lightweight models to distinguish between channel losses and collisions by learning the data trace from both simulation and real devices. Specifically, we first train a sophisticated teacher model based on extensive simulation datasets, whose knowledge is then transferred to a small student model that can be deployed on tiny device. The simulation-reality distillation is conducted via personalized trace from each client correspondingly, whose performance bound is analytically guaranteed. We have implemented our method on real testbed and show that the network access performance can be significantly improved, especially for sudden network variations.

Distributing the inference of convolutional neural network (CNN) to multiple mobile devices has been studied in recent years to achieve real-time inference without losing accuracy. However, how to map CNN to devices remains a challenge. On the one hand, scheduling the workload of state-of-the-art CNNs with multiple devices is NP-Hard because the structures of CNNs are directed acyclic graphs (DAG) rather than simple chains. On the other hand, distributing the inference workload suffers from expensive communication and unbalanced computation due to the wireless environment and heterogeneous devices. This paper presents PICO, a pipeline cooperation framework to accelerate the inference of versatile CNNs on diverse mobile devices. At its core, PICO features: (1) a generic graph partition algorithm that considers the characteristics of any given CNN and orchestrates it into a list of model pieces with suitable granularity, and (2) a many-to-many mapping algorithm that produces the best pipeline configuration for heterogeneous devices. In our experiment with$2 \sim 8$Raspberry-Pi devices, the throughput can be improved by$1.8 \sim 6.8 \times$under different CPU frequencies.

Privacy-preserving online multi-task assignment is a crucial aspect of spatial crowdsensing on untrusted platforms, where multiple real-time tasks are allocated to appropriate workers in a privacy-preserving manner. While existing schemes ensure the privacy of tasks and users, they seldom focus on minimizing the total moving distances for crowdsensing workers when assigning multiple tasks in real time, which adversely impacts the efficiency of online multi-task assignments. To address this issue, we propose POTA, the first privacy-preserving online multi-task assignment scheme with path planning that minimizes the total moving distances for crowdsensing workers without additional noise. POTA cryptographically implements the extended minimum-cost flow model, which models the encrypted data of workers and tasks in a graph and later produces optimized routing. With such a secure path-planning component, POTA reduces the total moving distances by$25.19\%-52.78\%$in the tested dataset compared with the state-of-the-art schemes with obfuscated path planning. Security analysis proves that POTA guarantees the confidentiality of sensitive data, a stronger security property than introducing obfuscation to sensitive data. Experimental evaluations on real-world datasets demonstrate the feasibility of POTA in terms of running time and its ability to achieve minimized total moving distances.

Single-image dehazing is a challenging task that requires both local details and global distribution. Existing methods face challenges in color imbalance and inconsistent details when predicting a haze-free image, because of their limitations in generalization from a specific setting (physics-based methods), capturing global information (CNN-based methods) and capturing detailed local information ( ViT-based methods). In response to these challenges, we propose a balanced low-cost hybrid network called LHNetV2 based on LHNetV1. The key insight of LHNetV2 is the effective fusion of different features, and a series of novel approaches is proposed to increase the running speed of the original LHNetV1. Firstly, building upon the Feature-aware Information Fusion method, we preserve the original Physical Embedding and Architecture Aggregation components in LHNetV1. Next, to overcome the speed bottleneck of LHNetV1, we enhance the calculation method of attention in the ViT sub-network and streamline the cross-stage interaction strategy in the CNN main-network. Finally, we introduce a dynamic adversarial loss function to bolster both the training stability and performance of LHNetV2. The experiments are extensively conducted on mainstream datasets, and the results demonstrate that LHNetV2 achieves the best balance between the performance and the running speed in single-image dehazing. The code is available at https://github.com/SHYuanBest/LHNet.

Age of Information, which emerged as a new metric to quantify the freshness of information, has attracted increasing interests recently. To optimize the system AoI, most existing works try to compute an efficient schedule from the point of data transmission. Unfortunately, at wireless-powered network edge, the charging schedule of the source nodes also needs to be decided besides data transmission. Thus, in this paper, we investigate the joint scheduling problem of data transmission and energy replenishment to optimize the maximum peak AoI at network edge with directional chargers. To the best of our knowledge, this is the first work that considers such two problems simultaneously. Firstly, the theoretical bounds of the maximum peak AoI with respect to the charging latency are derived. Secondly, for the minimum peak AoI scheduling problem with a single charger, an optimal scheduling algorithm is proposed to minimize the charging latency, and then a data transmission scheduling strategy is also given to optimize the maximum peak AoI. The proposed algorithm is proved to have a constant approximation ratio of up to 1.5. As for the scenario with multiple chargers, an approximate algorithm is also proposed to minimize the charging latency and the maximum peak AoI. Additionally, when the network bandwidth constraint is considered, the algorithm which considers the parallelism of the charging process and data transmission process is also proposed to reduce the latency and the maximum peak AoI. Finally, the theoretical analysis and simulation results verify that the proposed algorithms have high performance in terms of latency and AoI.

Edge network slicing promises better utilization of network resources by dynamically allocating resources on demand. However, addressing the imbalance between slice resources and user demands becomes challenging when complex user behaviors lead to bursty traffic within the edge network. Hence, we propose a comprehensive dynamic slice strategy with two coupled sub-strategies (i) bursty-sensitive slice resource coordination and (ii) proactive demand resource matching to find an optimal balance. For obtaining stable strategies, the edge network with bursty traffic is formulated as a bi-level Lyapunov optimization problem. Then we propose a resource allocation and request redirection (RA-RR) algorithm with polynomial complexity by introducing deep reinforcement learning to guarantee real-time. Specifically, two agents are trained to solve two sub-strategies, and the Lyapunov drift-plus-penalty function is used as the reward to keep queues stable. RA-RR is responsive to fluctuations in demand and realizes an efficient interaction of coupled decision-making. Moreover, a training method based on alternating optimization is designed to ensure convergence of the RA-RR algorithm. Experiments demonstrate that the proposal can maximize network revenue while ensuring the stability of slice services when edge traffic bursts, and has an average improvement of 20.4% compared with comparisons.

The emerging digital twin technique enhances the network management efficiency and provides comprehensive insights on network performance, through mapping physical objects to their digital twins. The user satisfaction on digital twin-enabled service relies on the freshness of digital twin data, which is measured by the Age of Information (AoI). Due to long service delays, the use of the remote cloud for delay-sensitive service provisioning faces serious challenges. Mobile Edge Computing (MEC), as an ideal paradigm for delay-sensitive services, is able to realize real-time data communication between physical objects and their digital twins at the network edge. However, the mobility of physical objects and dynamics of user query arrivals make seamless service provisioning in MEC become challenging. In this paper, we investigate dynamic digital twin placements for improving user service satisfaction in MEC environments, by introducing a novel metric to measure user service satisfaction based on the AoI concept and formulating two user service satisfaction enhancement problems: the static and dynamic utility maximization problems under static and dynamic digital twin placement schemes. To this end, we first formulate an Integer Linear Programming (ILP) solution to the static utility maximization problem when the problem size is small; otherwise, we propose a performance-guaranteed approximation algorithm. We then propose an online algorithm with a provable competitive ratio for the dynamic utility maximization problem, by considering dynamic user query services. Finally, we evaluate the performance of the proposed algorithms via simulations. Simulation results demonstrate that the proposed algorithms outperform the comparison baseline algorithms, improving the algorithm performance by at least$10.7\%$, compared to the baseline algorithms.

The Mobile Edge Computing (MEC) paradigm gives impetus to the vigorous advancement of the Internet of Things (IoT), through provisioning low-latency computing services at network edges. The emerging digital twin technique has been explosively growing in the IoT community, which bridges the gap between physical objects and their digital representations in an MEC network, enabling real-time monitoring and analysis, simulations on the dynamics of systems, accurate predictions on behaviours of objects, and optimization on network resource allocation. In this paper, we consider AoI-aware query services in an MEC network empowered by digital twin technology for diverse IoT applications. We aim to maximize the weighted sum of the accumulative freshness of query results measured by the Age of Information (AoI) and the total query service delay of admitted requests. To this end, we first formulate a novel minimization problem that explores nontrivial trade-offs between the two conflicting optimization objectives: the freshness of query results and service delays, and we show the NP-hardness of the problem. Then, we propose an approximation algorithm with a provable approximation ratio for the problem, at the expense of bounded computing capacity violations. We also develop a heuristic for the problem without any capacity violations. We finally evaluate the performance of the proposed algorithms via simulations. The simulation results demonstrate that the proposed algorithms are promising, and outperform the comparison benchmarks.

Federated Learning (FL) is to allow multiple clients to collaboratively train a model while keeping their data locally. However, existing FL approaches typically assume that the data in each client is static and fixed, which cannot account for incremental data with domain shift, leading to catastrophic forgetting on previous domains, particularly when clients are common edge devices that may lack enough storage to retain full samples of each domain. To tackle this challenge, we proposeFederatedDomain-IncrementalLearning viaSynergisticReplay (SR-FDIL), which alleviates catastrophic forgetting by coordinating all clients to cache samples and replay them. More specifically, when new data arrives, each client selects the cached samples based not only on their importance in the local dataset but also on their correlation with the global dataset. Moreover, to achieve a balance between learning new data and memorizing old data, we propose a novel client selection mechanism by jointly considering the importance of both old and new data. We conducted extensive experiments on several datasets of which the results demonstrate that SR-FDIL outperforms state-of-the-art methods by up to 4.05% in terms of average accuracy of all domains.

Foundation models (FMs), known as pre-trained models, have garnered significant interest in Industrial Internet due to their remarkable performance and robust generalization capabilities in downstream tasks. However, with the increasing requirements of computing infrastructure and data privacy protection for large foundation models, existing learning frameworks face challenges such as data privacy leakage, poor scalability, and deployment difficulties. To address these issues, this paper proposes a novel collaborative Transformer Block (TB)-wise training framework based on Federated Learning (FL), which consists of three stages: pre-training, graph regularization, and personalized training. To tackle the challenge of statistical heterogeneity in distributed data, we design a Graph Convolutional Network (GCN)-based update operator that captures local training representations. Besides, we conduct an analysis based on feature similarity to enhance the interpretability of our algorithm. We choose popular vision Transformer models for the experiments, extensive results demonstrate that our framework can jointly train multiple clients to build a foundation model while improving the single client's personalized performance. The proposed method outperforms state-of-the-art frameworks under various data distributions and system heterogeneity settings, highlighting its robust performance.

Age of Information (AoI) has emerged as a new metric to measure data freshness from the destination's perspective. To optimize the system AoI, most existing works focused on the point of scheduling of update transmissions. While at wireless-powered network edge, the source nodes can only transmit their updates after being charged ready, which means the system AoI is not only determined by the update transmission strategies, but also the charging strategies. Thus, in this paper, we investigate the first work to optimize the weighted peak AoI from the point of charging at wireless-powered network edge. Firstly, the problem of optimizing the weighted sum of average peak AoI with a directional charger is formulated, and then transformed to a charging time optimization problem with respect to the charging orientations and peak AoI, and an approximate algorithm is proposed to obtain the required charging time for each source node. Secondly, an age-based scheduling algorithm is proposed to compute the charging decisions and transmission decisions simultaneously, which can not only optimize the weighted sum of average peak AoI, but also guarantee the maximum peak AoI of each source node is bounded. The proposed algorithm is proved to have an approximation ratio of up to (1+$\varphi$), where$\varphi$is a small value related to the weight of each source node. When there exist multiple chargers, an approximate algorithm is also proposed to minimize the weighted sum of average peak AoI by scheduling the orientations of these chargers cooperatively. Finally, the extensive simulations demonstrate the high performance of the proposed algorithms in terms of peak AoI.

The advancement of information and telecommunication technology has resulted in lots of computing power service providers (CSPs) sharing resources. This effectively improves the utilization of computing resources and provides opportunities for users to find best network services. However, with the increase of similar or identical services, traditional service provision mechanisms become more complicated and face more challenges. To provide on-demand services for users and trusted service sharing platform for CSPs, we propose a blockchain-based distributed network service provision (DNSP) mechanism. Based on blockchain technology, we design a distributed network service (BBDNS) architecture and introduce trusted QoS model for users and profit model for CSPs. Meanwhile, we develop a bi-objective optimization problem, called BP-MATCH, to balance benefits between users and CSPs. Then, we define DNSP mechanism based on smart contracts. Furthermore, to solve BP-MATCH, we design Kuhn-Munkres based service matching algorithm (KM-SMA) and ant colony optimization-based service matching algorithm (ACO-SMA). Finally, we conduct a simulation experiment based on the generated dataset. Simulation results show that the proposed algorithm can obtain the optimal service matching decision under certain conditions, and DNSP mechanism can guarantee the efficiency of service decisions while providing trusted distributed service.

Recently, federated learning (FL) has received intensive research because of its ability in preserving data privacy for scattered clients to collaboratively train machine learning models. Decentralized federated learning (DFL) is upgraded from FL which allows clients to aggregate model parameters with their neighbours directly. DFL is particularly feasible for dynamic systems, in which the neighbour set of each client is dynamic. However, due to the restrictions of client trajectories and communication distances, it is hard for individual clients to sufficiently exchange models with others, resulting in poor model accuracy. To address this challenge, we propose the DFL-DMS (DFL with Diversified Model Sources) algorithm to diversify sources for model aggregation, and improve model utility. Specifically, models exchanged between DFL-DMS clients are jointly determined by their staleness scores and the bandwidth constraint. An asynchronous learning mode is adopted so that DFL-DMS clients can temporarily store and relay fresh models collected from different client sources to accelerate the propagation of rare models. A state vector is maintained to track the contribution weight of each source to its model aggregation, and an entropy based metric (EBM) is optimized by clients in a distributed manner. Finally, the superiority of DFL-DMS is evaluated by extensive experiments (with MNIST and CIFAR-10 datasets) which demonstrate that DFL-DMS can accelerate the convergence of DFL and improve the model accuracy significantly compared with the state-of-the-art baselines.

In recent years, the rapid growth of Internet of Things (IoT) devices and applications has posed significant challenges for existing Mobile Edge Computing (MEC) architectures. The inherent latency uncertainties in MEC architectures make it difficult to support latency-sensitive applications such as autonomous vehicles. Additionally, the increasing number of connected devices has led to substantial challenges in terms of limited throughput for MEC servers. With the recent advancements in programmable network hardware, such as SmartNICs and programmable switches, the Network-based Computing (NBC) paradigm has gained widespread attention. Leveraging line-rate processing capabilities, NBC offers a promising solution for high throughput and low latency processing. This paper aims to explore the potential benefits and challenges of incorporating NBC into existing MEC architectures. The feasibility of our proposed architecture is evaluated using two use cases, Linear Quadratic Regulator (LQR) control and Complex Event Processing (CEP), demonstrating significant improvements in latency performance.

The flying ad hoc networks (FANETs) have attracted a large amount of attention from both academia and industry. Benefiting from the flexibility, the FANETs have been widely deployed in various scenarios, ranging from agricultural production to emergency rescue. However, in FANETs, the mobility of unmanned aerial vehicles (UAVs) has led to critical challenges for the stability of communications. Especially, the routing flooding mechanism extremely limits the scalability of FANET. To overcome these technical challenges, constructing a hierarchy and clustering structure in FANETs is considered a promising solution. In this paper, we propose the fission spectral clustering (FSC) strategy for UAV swarm networks. We model the UAV clustering problem as a graph cut problem. The time-sequential attributes weight of nodes and edges will be input to the FSC algorithm. Then, it will construct the Laplace matrix and calculate the first k-th eigenvectors of it. We apply the K-Means algorithm into this feature space to cut the graph by clustering the eigenvectors. Each cluster will constantly fission with this strategy until it satisfies the size and structure constraints in the UAV clusters. Some simulations are implemented to evaluate our proposed algorithm in comparison to the other state-of-the-art solutions.

This paper investigates an intelligent reflecting surfaces (IRS) aided anti-jamming communication strategy in the integrated terrestrial-satellite network (ITSN), where the IRS is exploited to mitigate jamming interference and enhance the integrated system communication performance. In such a network, the terrestrial network and satellite network are co-existing with a spectrum-sharing scheme in the presence of a multi-antenna jammer. We aim at maximizing the weighted sum rate (WSR) of all users by jointly optimizing the terrestrial beamformers and IRS phase shifts while considering the signal-to-interference-plus-noise ratio (SINR) requirements of legitimate users. Different from the non-convex optimization techniques utilized in the IRS-related problem, a novel optimization-driven deep reinforcement learning (DRL) algorithm is proposed, which leverages both the robustness of model-free learning approaches and the efficiency of model-based optimization methods. In the optimization module of the proposed algorithm, we analyze the smart jammer under the unknown jamming model and derive a lower bound of the anti-jamming uncertainty, such that the IRS-aided anti-jamming problem can be solved by alteration method with second-order cone programming (SOCP) algorithm and semidefinite relaxation (SDR) technique. Simulation results demonstrate that the IRS can enhance the anti-jamming performance efficiently, and the proposed optimization-driven DRL algorithm can improve both the learning rate and the system performance compared with existing solutions.

To satisfy the requirements of content distribution in computation-intensive and delay-sensitive services, this paper presents a novel joint task offloading and content caching (JTOCC) scheme in multi-cell multi-carrier non-orthogonal multiple-access (MCMC-NOMA)-assisted cloud-edge-terminal cooperation networks. Based on queuing theory, we formulate a delay minimization model that aggregates users’ requests to reduce repeated content delivery. To minimize network latency, the model is decomposed into three subproblems: task offloading, user clustering and communication resource allocation, and cache state updating. In each slot, the task offloading subproblem is solved utilizing deep reinforcement learning (DRL) under a resource-constrained cloud-edge-terminal setting. During a transition between slots, mobile terminals are grouped using K-means-based user clustering, and the allocations of the subchannels and transmit power are optimized utilizing matching theory and successive convex approximation (SCA), respectively. Contents cached at the network nodes are updated, according to long-short-term memory (LSTM)-based predicted popularity. Simulations show that the proposed JTOCC model achieves lower-delay content distribution than its existing counterparts in cloud-edge-terminal cooperation environments, and converges fast in heterogeneous networks.

The growth of intelligent vehicular services, like augmented reality (AR) road simulation, underscores the need for rapid, multi-modal content generation. Generative artificial intelligence (GAI) models, known for their swift production of diverse artificial intelligence-generated content (AIGC), stand out as a prime solution. However, integrating cloud-centric GAI models into vehicular networks is fraught with challenges. Notably, to offer specialized generative edge intelligence (EI) and boost vehicular AIGC, GAI models need to tap into user data and utilize significant computation resources. Moreover, their deployment across vehicular networks is essential for proximity-based distributed inferences. Yet, edge devices are resource-limited, and data sharing can raise safety and privacy concerns. Addressing these challenges, this paper introduces GAI-IoV, an EI-enabled GAI framework facilitated through the cooperation between road-side units (RSUs) and vehicles. Subsequently, we propose the workflow for collaborative fine-tuning and distributed inference. On this basis, two pivotal vehicle-centric problems are then formulated: computation and communication resource allocation for federated fine-tuning (FFT) to optimize time and energy cost, and splitting strategy of shared and local inferences to optimize inference latency and content-generation capability. To solve these optimizations, we introduce a self-adaptive global best harmony search (SGHS) algorithm for resource allocation and a backward induction method for determining inference splitting strategy. Our experiments based on the Stable Diffusion v1-4 model vouch for a superior fine-tuning and inference capabilities of GAI-IoV. Furthermore, simulations underscore its resource utilization and distributed inference efficiency in dynamic vehicular scenarios.

Even the artificial intelligence (AI) has been widely used and significantly changed our life, deploying the large AI models on resource limited edge devices directly is not appropriate. Thus, the model split inference is proposed to improve the performance of edge intelligence (EI), in which the AI model is divided into different sub-models and the resource-intensive sub-model is offloaded to edge server wirelessly for reducing resource requirements and inference latency. Unfortunately, with the sharp increasing of edge devices, the shortage of spectrum resource in edge network becomes seriously in recent years, which limits the performance improvement of EI. Refer to the NOMA-based edge computing (EC), integrating non-orthogonal multiple access (NOMA) technology with split inference in EI is attractive. However, the NOMA-based communication aspect and the influence of intermediate data transmission fail to be considered properly in model split inference of EI in previous works, and the sophistication in resource allocation caused by NOMA scheme makes it further complicated. Thus, the Effective Communication and Computing resource allocation algorithm is proposed in this paper for accelerating the split inference in NOMA-based EI, shorted as ECC. Specifically, the ECC takes the energy consumption and the inference latency into account to find the optimal model split strategy and resource allocation strategy (subchannel, transmission power, computing resource). Since the minimum inference delay and energy consumption cannot be satisfied simultaneously, the gradient descent (GD) based algorithm is adopted to find the optimal tradeoff between them. Moreover, the loop iteration GD approach (Li-GD) is developed to reduce the complexity of the GD algorithm caused by parameter discretization. The key idea of Li-GD is that: the initial value of the$i\mathrm {th}$layer’s GD procedure is selected from the optimal results of the former$(i-1)$layers’ GD procedure whose intermediate data size is the closest to$i\mathrm {th}$layer. Additionally, the properties of the proposed algorithms are investigated, including convergence, complexity, and approximation error. The experimental results demonstrate that the performance of ECC is much better than that of the previous studies.

Zero-shot learning (ZSL) is an extreme case of transfer learning that aims to recognize samples (e.g., images) of unseen classes relying on a train-set covering only seen classes and a set of auxiliary knowledge (e.g., semantic descriptors). Existing methods usually resort to constructing a visual-to-semantics mapping based on features extracted from each whole sample. However, since the visual and semantic spaces are inherently independent and may exist in different manifolds, these methods may easily suffer from the domain bias problem due to the knowledge transfer from seen to unseen classes. Unlike existing works, this paper investigates the fine-grained ZSL from a novel perspective of sample-level graph. Specifically, we decompose an input into several fine-grained elements and construct a graph structure per sample to measure and utilize element-granularity relations within each sample. Taking advantage of recently developed graph neural networks (GNNs), we formulate the ZSL problem to a graph-to-semantics mapping task, which can better exploit element-semantics correlation and local sub-structural information in samples. Experimental results on the widely used benchmark datasets demonstrate that the proposed method can mitigate the domain bias problem and achieve competitive performance against other representative methods.

Recent studies have demonstrated that local training data in Federated Learning can be recovered from gradients, which are called gradient inversion attacks. These attacks display powerful effects on either computer vision or natural language processing tasks. As it is known that there are certain correlations between multi-modality data, we argue that the threat of such attacks combined with Multi-modal Learning may cause more severe effects. Different modalities may communicate through gradients to provide richer information for the attackers, thus improving the strength and efficiency of the gradient inversion attacks. In this paper, we propose the Mutual Gradient Inversion Attack (MGIA), by utilizing the shared labels between image and text modalities combined with the idea of knowledge distillation. Our experimental results show that MGIA achieves the best quality of both modality data and label recoveries in comparison with other methods. In the meanwhile, MGIA verifies that multi-modality gradient inversion attacks are more likely to disclose private information than the existing single-modality attacks.

Real-time video perception tasks are often challenging over the resource-constrained edge devices due to the concerns of accuracy drop and hardware overhead, where saving computations is the key to performance improvement. Existing methods either rely on domain-specific neural chips or priorly searched models, which require specialized optimization according to different task properties. In this work, we propose a general and task-independent Patch Automatic Skip Scheme (PASS), a novel end-to-end learning pipeline to support diverse video perception settings by decoupling acceleration and tasks. The gist is to capture the temporal similarity across video frames and skip the redundant computations at patch level, where the patch is a non-overlapping square block in visual. PASS equips each convolution layer with a learnable gate to selectively determine which patches could be safely skipped without degrading model accuracy. As to each layer, a desired gate needs to make flexible skip decisions based on intermediate features without any annotations, which cannot be achieved by conventional supervised learning paradigm. To address this challenge, we are the first to construct a tough self-supervisory procedure for optimizing these gates, which learns to extract contrastive representation, i.e., distinguishing similarity and difference, from frame sequence. These high-capacity gates can serve as a plug-and-play module for convolutional neural network (CNN) backbones to implement patch-skippable architectures, and automatically generate proper skip strategy to accelerate different video-based downstream tasks, e.g., outperforming the state-of-the-art MobileHumanPose (MHP) in 3D pose estimation and FairMOT in multiple object tracking, by up to 9.43 times and 12.19 times speedups, respectively. By directly processing the raw data of frames, PASS can generalize to real-time video streams on commodity edge devices, e.g., NVIDIA Jetson Nano, with efficient performance in realistic deployment.

Multimodal learning (MML) aims to jointly exploit the common priors of different modalities to compensate for their inherent limitations. However, existing MML methods often optimize a uniform objective for different modalities, leading to the notorious “modality imbalance” problem and counterproductive MML performance. To address the problem, some existing methods modulate the learning pace based on the fused modality, which is dominated by the better modality and eventually results in a limited improvement on the worse modal. To better exploit the features of multimodal, we propose Prototypical Modality Rebalance (PMR) to perform stimulation on the particular slow-learning modality without interference from other modalities. Specifically, we introduce the prototypes that represent general features for each class, to build the non-parametric classifiers for uni-modal performance evaluation. Then, we try to accelerate the slow-learning modality by enhancing its clustering toward prototypes. Furthermore, to alleviate the suppression from the dominant modality, we introduce a prototype-based entropy regularization term during the early training stage to prevent premature convergence. Besides, our method only relies on the representations of each modality and without restrictions from model structures and fusion methods, making it with great application potential for various scenarios. The source code is available here11https://github.com/fanyunfeng-bit/Modal-Imbalance-PMR.

Recent studies usually approach multi-label zeroshot learning (MLZSL) with visual-semantic mapping on spatial-class correlation, which can be computationally costly, and worse still, fails to capture fine-grained classspecific semantics. We observe that different channels may usually have different sensitivities on classes, which can correspond to specific semantics. Such an intrinsic channelclass correlation suggests a potential alternative for the more accurate and class-harmonious feature representations. In this paper, our interest is to fully explore the power of channel-class correlation as the unique base for MLZSL. Specifically, we propose a light yet efficient Multi-Label Multi-Layer Perceptron-based Encoder, dubbed (ML)2P-Encoder, to extract and preserve channel-wise semantics. We reorganize the generated feature maps into several groups, of which each of them can be trained independently with (ML)2P-Encoder. On top of that, a global groupwise attention module is further designed to build the multilabel specific class relationships among different classes, which eventually fulfills a novel Channel-Class Correlation MLZSL framework (C3-MLZSL)11Released code:github.com/simonzmliu/cvpr23_mlzsl. Extensive experiments on large-scale MLZSL benchmarks including NUS-WIDE and Open-Images-V4 demonstrate the superiority of our model against other representative state-of-the-art models.

Compositional Zero-Shot Learning (CZSL) aims to recognize novel concepts formed by known states and objects during training. Existing methods either learn the combined state-object representation, challenging the generalization of unseen compositions, or design two classifiers to identify state and object separately from image features, ignoring the intrinsic relationship between them. To jointly eliminate the above issues and construct a more robust CZSL system, we propose a novel framework termed Decomposed Fusion with Soft Prompt (DFSP)11Code is available at: https://github.corn/Forest-art/DFSP.git, by involving vision-language models (VLMs)for unseen composition recognition. Specifically, DFSP constructs a vector combination of learnable soft prompts with state and object to establish the joint representation of them. In addition, a cross-modal decomposed fusion module is designed between the language and image branches, which decomposes state and object among language features instead of image features. Notably, being fused with the decomposed features, the image features can be more expressive for learning the relationship with states and objects, respectively, to improve the response of unseen compositions in the pair space, hence narrowing the domain gap between seen and unseen sets. Experimental results on three challenging benchmarks demonstrate that our approach significantly outperforms other state-of-the-art methods by large margins.

Recently, diverse emerging scenarios have precipitated a substantial surge in the variety of devices and applications, which has consequently imposed more stringent demands on Quality of Service (QoS) prerequisites. As a burgeoning emerging network measurement method, In-band network telemetry (INT), can provide detailed metrics for QoS by obtaining fine-grained network status information. However, INT only outlines device-level operations, which fails to provide an entire network view for monitoring. To address this, INT orchestration based on network topology and application requirements to achieve network-level monitoring is necessary. In this paper, we propose an INT orchestration model that efficiently measures the entire network while minimizing measuring overhead. The model outputs the probe path and collects requirements for the devices it passes through. Our method effectively reduces network bandwidth consumption caused by INT process and ensures telemetry items remain fresh. Experiment results support the effectiveness of our approach.

Most existing neural network pruning methods hand-crafted their importance criteria and structures to prune. This constructs heavy and unintended dependencies on heuristics and expert experience for both the objective and the parameters of the pruning approach. In this paper, we try to solve this problem by introducing a principled and unified framework based on Information Bottleneck (IB) theory, which further guides us to an automatic pruning approach. Specifically, we first formulate the channel pruning problem from an IB perspective, and then implement the IB principle by solving a Hilbert-Schmidt Independence Criterion (HSIC) Lasso problem under certain conditions. Based on the theoretical guidance, we then provide an automatic pruning scheme by searching for global penalty coefficients. Verified by extensive experiments, our method yields state-of-the-art performance on various benchmark networks and datasets. For example, with VGG-16, we achieve a 60%-FLOPs reduction by removing 76% of the parameters, with an improvement of 0.40% in top-1 accuracy on CIFAR-10. With ResNet-50, we achieve a 56%-FLOPs reduction by removing 50% of the parameters, with a small loss of 0.08% in the top-1 accuracy on ImageNet. The code is available at https://github.com/sunggo/APIB.

Cross-silo federated learning (CFL) is a distributed learning paradigm that allows organizations (e.g., financial or medical entities) to train a global model on siloed data. Recent studies on mechanisms designed for CFL, however, rarely jointly consider the potential inter-organizational competition and the lack of credibility between organizations, which may discourage organizational participation. In this paper, we investigate the problem of inter-organizational competition and credibility assurance. We propose a distributed trading mechanism, called$TradeFL$, to incentivize organizations to contribute data and computational resources through mutual trading among organizations. Technically, TradeFL characterizes the competition among organizations and compensates for their damage incurred by competition. TradeFL runs on distributed organizations and provides credibility guarantees for compensation through a customized smart contract11Illustration of the prototype: https://github.com/user10963.. We prove that the interaction among organizations that contribute resources to maximize personal payoffs is a weighted potential game. Then, we propose a centralized algorithm and a distributed algorithm to determine the optimal resource contribution. Simulation results and evaluations based on real-world datasets demonstrate that our scheme achieves higher social welfare, increases the amount of contributed data by up to 64%, and improves the accuracy of the global model by at most 23.2%.

Adolescent idiopathic scoliosis is becoming a common spinal disorder among adolescents. The traditional methods of scoliosis screening are labor-intensive and can result in unnecessary referrals and radiological exposure for adolescents due to their low positive predictive value. For early screening of scoliosis and abnormal posture, a mobile-based cost-free, accurate and radiation-free scoliosis screening system is proposed in this paper. We establish a database with labeled 2D unclothed back images and corresponding whole-spine standing posterior-anterior X-ray images, and innovatively propose a new network topology of the 2D back image to localize the back landmarks. With only an unclothed back image, this system can automatically classify normal, abnormal posture and scoliosis with an overall classification accuracy of 88.1%. This system has the potential to overcome the time and space limitations of conventional screening for scoliosis and abnormal posture.

Compared with full client participation, partial client participation is a more practical scenario in federated learning, but it may amplify some challenges in federated learning, such as data heterogeneity. The lack of inactive clients’ updates in partial client participation makes it more likely for the model aggregation to deviate from the aggregation based on full client participation. Training with large batches on individual clients is proposed to address data heterogeneity in general, but their effectiveness under partial client participation is not clear. Motivated by these challenges, we propose to develop a novel federated learning framework, referred to as FedAMD, for partial client participation. The core idea is anchor sampling, which separates partial participants into anchor and miner groups. Each client in the anchor group aims at the local bullseye with the gradient computation using a large batch. Guided by the bullseyes, clients in the miner group steer multiple near-optimal local updates using small batches and update the global model. By integrating the results of the two groups, FedAMD is able to accelerate the training process and improve the model performance. Measured by $\epsilon$-approximation and compared to the state-of-the-art methods, FedAMD achieves the convergence by up to $O(1/\epsilon)$ fewer communication rounds under non-convex objectives. Empirical studies on real-world datasets validate the effectiveness of FedAMD and demonstrate the superiority of the proposed algorithm: Not only does it considerably save computation and communication costs, but also the test accuracy significantly improves.

Federated Semi-supervised Learning (FedSSL) has emerged as a new paradigm for allowing distributed clients to collaboratively train a machine learning model over scarce labeled data and abundant unlabeled data. However, existing works for FedSSL rely on a closed-world assumption that all local training data and global testing data are from seen classes observed in the labeled dataset. It is crucial to go one step further: adapting FL models to an open-world setting, where unseen classes exist in the unlabeled data. In this paper, we propose a novel Federatedopen-world Semi-Supervised Learning (FedoSSL) framework, which can solve the key challenge in distributed and open-world settings, i.e., the biased training process for heterogeneously distributed unseen classes. Specifically, since the advent of a certain unseen class depends on a client basis, the locally unseen classes (exist in multiple clients) are likely to receive differentiated superior aggregation effects than the globally unseen classes (exist only in one client). We adopt an uncertainty-aware suppressed loss to alleviate the biased training between locally unseen and globally unseen classes. Besides, we enable a calibration module supplementary to the global aggregation to avoid potential conflicting knowledge transfer caused by inconsistent data distribution among different clients. The proposed FedoSSL can be easily adapted to state-of-the-art FL methods, which is also validated via extensive experiments on benchmarks and real-world datasets (CIFAR-10, CIFAR-100 and CINIC-10).

Federated learning (FL) is an emerging technique that trains massive and geographically distributed edge data while maintaining privacy. However, FL has inherent challenges in terms of fairness and computational efficiency due to the rising heterogeneity of edges, and thus usually results in sub-optimal performance in recent state-of-the-art (SOTA) solutions. In this paper, we propose a Customized Federated Learning (CFL) system to eliminate FL heterogeneity from multiple dimensions. Specifically, CFL tailors personalized models from the specially designed global model for each client jointly guided by an online trained model-search helper and a novel aggregation algorithm. Extensive experiments demonstrate that CFL has full-stack advantages for both FL training and edge reasoning and significantly improves the SOTA performance w.r.t. model accuracy (up to 7.2% in the non-heterogeneous environment and up to 21.8% in the heterogeneous environment), efficiency, and FL fairness.

By combing edge computing and parallel computing, distributed edge computing has emerged as a new paradigm to accelerate computation at the edge. Considering the parallelism of both computation and communication, the problem of Minimum Latency joint Communication and Computation Scheduling (MLCCS) is studied recently. However, existing works have rigid assumptions that the communication time of each device is fixed and the workload can be split arbitrarily small. Aiming at making the work more practical and general, the MLCCS problem without the above assumptions is studied in this paper. Firstly, the MLCCS problem under a general model is formulated and proved to be NP-hard. Secondly, a pyramid-based computing model is proposed to consider the parallelism of communication and computation jointly, which has an approximation ratio of 1 + δ, where δ is related to devices’ communication rates. An interesting property under such computing model is identified and proved, i.e., the optimal latency can be obtained under arbitrary scheduling order when all the devices share the same communication rate. When the devices own different communication rates, the optimal scheduling order is also obtained. Additionally, when the workload cannot be split arbitrarily, an approximation algorithm with ratio of at most 2 (1 + δ) is proposed. Finally, the theoretical analysis and simulation results verify that the proposed algorithm has high performance in terms of latency. Two testbed experiments are also conducted, which show that the proposed method outperforms the existing methods, reducing the latency by up to 29.2% in real-world applications.

Sharding scales throughput by splitting blockchain nodes into parallel groups. However, different shards’ independent and random scheduling for cross-shard transactions results in numerous conflicts and aborts, since cross-shard transactions from different shards may access the same account. A deterministic ordering can eliminate conflicts by determining a global order for transactions before processing, as proved in the database field. Unfortunately, due to the intertwining of the Byzantine environment and information isolation among shards, there is no trusted party able to predetermine such an order for cross-shard transactions. To tackle this challenge, this paper proposes Prophet, a conflict-free sharding blockchain based on Byzantine-tolerant deterministic ordering. It first depends on untrusted self-organizing coalitions of nodes from different shards to pre-execute cross-shard transactions for prerequisite information about ordering. It then determines a trusted global order based on stateless ordering and post-verification for pre-executed results, through shard cooperation. Following the order, the shards thus orderly execute and commit transactions without conflicts. Prophet orchestrates the pre-execution, ordering, and execution processes in the sharding consensus for minimal overhead. We rigorously prove the determinism and serializability of transactions under the Byzantine and sharded environment. An evaluation of our prototype shows that Prophet improves the throughput by 3.11× and achieves nearly no aborts on 1 million Ethereum transactions compared with state-of-the-art sharding.

Age of Information, which emerged as a new performance metric to quantify the data freshness, has drawn increasing interests recently. At wireless-powered network edge, the source nodes can only transmit their updates after being charged ready, which means the system AoI is not only determined by the update transmission strategies, but also the charging strategies. However, the existing works either only focused on the point of scheduling of update transmissions or have a rigid assumption that only one source node can be charged per time. Aiming at making the work more practical and general, we investigate the average$A$oI optimization problem at wireless-powered network edge without such limitations in this paper. Firstly, the lower bound of the weighted sum of average$A$oI of the whole network with a directional charger is analyzed, which is proved to be related to nodes' maximum transmitting interval and the charging strategy. An optimal charging time allocation algorithm is proposed to obtain the maximum transmitting interval of each source node by considering the overlapped areas of different charging orientations. After then, an AoI-aware periodical charging scheduling algorithm is proposed, which can obtain a periodical charging schedule including a charging period$T$and the charging orientation assigned to each time slot within$T$, while the average AoI is bounded. The proposed algorithm is proved to have an approximation ratio of up to 1.5625. Finally, the extensive simulations demonstrate the high performance of the proposed algorithm in terms of AoI.

The recent success of pre-trained language models (PLMs) such as BERT has resulted in the development of various beneficial database middlewares, including natural language query interfaces and entity matching. This shift has been greatly facilitated by the extensive external knowledge of PLMs. However, as PLMs are often provided by untrusted third parties, their lack of standardization and regulation poses significant security risks that have yet to be fully explored. This paper investigates the security threats posed by malicious PLMs to these emerging database middleware. We specifically propose a novel type of Trojan attack, where a maliciously designed PLM causes unexpected behavior in the database middleware. These Trojan attacks possess the following characteristics: (1) Triggerability: The Trojan-infected database middleware will function normally with normal input, but will likely malfunction when triggered by the attacker. (2) Imperceptibility: There is no need for noticeable modification of the input to trigger the Trojan. (3) Generalizability: The Trojan is capable of targeting a variety of downstream tasks, not just one specific task. We thoroughly evaluate the impact of these Trojan attacks through experiments and analyze potential countermeasures and their limitations. Our findings could aid in the creation of stronger mechanisms for the implementation of PLMs in database middleware.

Single image dehazing is a challenging task that requires both local detail and global distribution, and can be applied to various scenarios. However, physics-based dehazing algorithms perform well only in specific settings, while CNN-based algorithms struggle with capturing global information, and ViT-based approaches suffer from inadequate representation of local details. The shortcomings of the above three types of methods lead to issues such as imbalanced colors and incoherent details in the predicted haze-free image. To address these challenges, we propose a new Low-cost Hybrid Network called LHNet. The key insight of LHNet is the effective hybrid of different features, which can achieve better information fusion in the form of feature awareness at the cost of few parameters. This fusion approach narrows the gap between different features and enables LHNet to autonomously choose the fusion granularity to maximize the utilization of prior, local and global information. Extensive experiments are performed on the mainstream dehazing datasets, and the results show that LHNet achieves state-of-the-art performance in single image dehazing. By adopting our fusion approach, a better dehazing effect can be achieved than with other dehazing algorithms with more parameters, even when only CNN and ViT are used. The code is available at https://github.com/SHYuanBest/LHNet.

Generative models produce unbounded outputs, necessitating the use of refusal techniques to confine their output space. Employing generative refusals is crucial in upholding the ethical and copyright integrity of synthesized content, particularly when working with widely adopted diffusion models. "Concept negation'' presents a promising paradigm to achieve generative refusals, as it effectively defines and governs the model's output space based on concepts, utilizing natural language interfaces that are readily comprehensible to humans. However, despite the valuable contributions of prior research to the field of concept negation, it still suffers from significant limitations. The existing concept negation methods, which operate based on the composition of score or noise predictions from the diffusion process, are limited to independent concepts (e.g., ``a blonde girl`` without ``glasses``) and fail to consider the interconnected nature of concepts in reality (e.g., ``Mickey mouse eats ice cream`` without ``Disney characters``). Keeping the limitations in mind, we propose a novel framework, called $ProtoRe$, to improve the flexibility of concept negation via test-time negative concept identification along with purification in the feature space. $ProtoRe$ works by incorporating CLIP's language-contrastive knowledge to identify the prototype of negative concepts, extract the negative features from outputs using the prototype as a prompt, and further refine the attention maps by retrieving negative features. Our evaluation on multiple benchmarks shows that $ProtoRe$ outperforms state-of-the-art methods under various settings, in terms of the effectiveness of purification and the fidelity of generative images.

Test-time adaptation (TTA) is a special and practical setting in unsupervised domain adaptation, which allows a pre-trained model in a source domain to adapt to unlabeled test data in another target domain. To avoid the computation-intensive backbone fine-tuning process, the zero-shot generalization potentials of the emerging pre-trained vision-language models (e.g., CLIP, CoOp) are leveraged to only tune the run-time prompt for unseen test domains. However, existing solutions have yet to fully exploit the representation capabilities of pre-trained models as they only focus on the entropy-based optimization and the performance is far below the supervised prompt adaptation methods, e.g., CoOp. In this paper, we propose SwapPrompt, a novel framework that can effectively leverage the self-supervised contrastive learning to facilitate the test-time prompt adaptation. SwapPrompt employs a dual prompts paradigm, i.e., an online prompt and a target prompt that averaged from the online prompt to retain historical information. In addition, SwapPrompt applies a swapped prediction mechanism, which takes advantage of the representation capabilities of pre-trained models to enhance the online prompt via contrastive learning. Specifically, we use the online prompt together with an augmented view of the input image to predict the class assignment generated by the target prompt together with an alternative augmented view of the same image. The proposed SwapPrompt can be easily deployed on vision-language models without additional requirement, and experimental results show that it achieves state-of-the-art test-time adaptation performance on ImageNet and nine other datasets. It is also shown that SwapPrompt can even achieve comparable performance with supervised prompt adaptation methods.

Pre-trained vision-language models like CLIP show great potential in learning representations that capture latent characteristics of users. A recently proposed method called Contextual Optimization (CoOp) introduces the concept of training prompt for adapting pre-trained vision-language models. Given the lightweight nature of this method, researchers have migrated the paradigm from centralized to decentralized system to innovate the collaborative training framework of Federated Learning (FL). However, current prompt training in FL mainly focuses on modeling user consensus and lacks the adaptation to user characteristics, leaving the personalization of prompt largely under-explored. Researches over the past few years have applied personalized FL (pFL) approaches to customizing models for heterogeneous users. Unfortunately, we find that with the variation of modality and training behavior, directly applying the pFL methods to prompt training leads to insufficient personalization and performance. To bridge the gap, we present pFedPrompt, which leverages the unique advantage of multimodality in vision-language models by learning user consensus from linguistic space and adapting to user characteristics in visual space in a non-parametric manner. Through this dual collaboration, the learned prompt will be fully personalized and aligned to the user’s local characteristics. We conduct extensive experiments across various datasets under the FL setting with statistical heterogeneity. The results demonstrate the superiority of our pFedPrompt against the alternative approaches with robust performance.

The traditional centralized resource scheduling method leads to trust issues among multiple subjects carrying microservices. At the same time, in the process of service provision, single-point failure problems also occur from time to time. In order to realize the trusted provision of services, we build a blockchain-based distributed intelligent service trusted provision architecture, which uses smart contracts to realize the on-chain registration of resource information and automatic orchestration of microservices. In order to break through the bottleneck of blockchain efficiency and improve scalability, we use sharding technology to expand the blockchain. And the Raft-practical Byzantine fault-tolerance two-level consensus mechanism combining the Boneh–Lynn–Sacham (BLS) threshold signature (B-RBFT) is designed for blockchain sharding, which greatly improves throughput and reduces consensus delay while taking security into account. To meet higher Quality-of-Service (QoS) requirements, we design the microservice orchestration algorithm based on the improved double deep$Q$network (DDQN) to support microservice deployment and migration. In particular, to make the neural network converge faster, we improve the traditional DDQN framework by using double replay buffers and weighted target values. Simulation results show that our proposed algorithm has advantages in convergence speed, resource usage cost, delay, and load balancing.

The proliferation of users and data traffic poses substantial pressure on resource management in the Internet of Things (IoT). In addition to beneficially allocating scarce network resources, it also needs to meet differentiated users’ Quality-of-Service (QoS) requirements, such as low delay, high security, etc. The distributed management architecture of blockchain and its inherent security features bring inspiration to resource management in the IoT. In this article, we propose a blockchain-enabled resource orchestration scheme for IoT by deep reinforcement learning (DRL), where the IoT edge server and the end user can reach a consensus on the allocation of network resources based on blockchain theory. Moreover, relying on the policy network, the intelligent agent can be trained by these resource attributes to fully perceive the change of the network’s state and hence make dynamic resource allocation decisions. Finally, simulation results show that the proposed resource orchestration scheme has good performance in comparison to other security resource allocation algorithms. The average revenue, the user request acceptance rate, and the profitability are increased by an average of 8.5%, 1.8%, and 11.9%, respectively, compared with other algorithms.

Sparse sensor placement strategies are applied to reconstruct a region’s full-state data conditioned to a limited number of sensors; particularly, crucial to ocean monitoring systems. In maritime systems, existing sparse sensor placement methods mainly consider the reconstruction error of data or rely on specific requirements. Considering how sensors acquire essential information for monitoring systems, the utilization of entropy from information theory becomes quite interesting. In this article, we show that entropy measurements on different quantities of information are sensitive to indicate the border areas, thus requiring a balance between the number of sensors needed and the amount of information collected by them in coastal areas. Due to such, we propose: 1) a novel sparse sensor placement strategy based on entropy, where the entropy measurements in temporal dimension are utilized for sample selection, so portions of samples selected are utilized for training data, significantly improving the training efficiency without sacrificing accuracy of subsequent data reconstruction. In the proposed strategy, 2) we use orthogonal triangle decomposition from linear algebra where a low-cost sensor is employed as pivot and in terms of spatial dimension, the entropy of each location is adopted as entropy weight to reconstruct full state data. Additionally, 3) the strategy employs a greedy algorithm of weighted column pivoting for the orthogonal triangle decomposition, which is designed to suit yet effectively seek additional information and minimal reconstruction error in each iteration processing step. Experimental results using sea surface temperature (SST) data show that the proposed strategy outperforms existing methods, acquiring more information, ensuring higher efficiency, and reducing costs while minimizing reconstruction errors.

6G edge networks strive to offer ubiquitous intelligent services, requiring a greater emphasis on network stability and reliability. However, current networks present a low automation degree of the operation, administration and maintenance process. Consequently, active service migration away from abnormal network nodes and links, as well as automatic and transparent service recovery from sudden anomalies, become challenging tasks. These conditions underscore the urgency for an innovative service self-healing mechanism for 6G edge networks. Digital twin (DT) technology uses modeling to represent physical entities, thereby facilitating lifecycle management. However, the application of DT technology in networks is still a burgeoning field of study. In this paper, we explore the DT-driven service self-healing mechanism in 6G edge networks. Initially, we design a DT-based architecture for service self-healing. Subsequently, we construct a performance prediction mechanism leveraging graph neural networks (GNNs) to devise an efficient prediction model, which aims to accurately infer network performance and promptly detect abnormal network conditions. To maintain fine-grained service stability amidst potential network anomalies, we propose a DT-driven service redeployment mechanism enhanced by GNNs. Comprehensive experimental results reveal that our proposed mechanism can accurately predict flow-level delays and identify abnormal links and nodes. Furthermore, the DT-driven service redeployment mechanism effectively reduces service delay and enhances network load balance.

Blockchain databases have attracted widespread attention but suffer from poor scalability due to underlying non-scalable blockchains. While blockchain sharding is necessary for a scalable blockchain database, it poses a new challenge named on-chain cross-shard database services. Each cross-shard database service (e.g., cross-shard queries or inter-shard load balancing) involves massive cross-shard data exchanges, while the existing cross-shard mechanisms need to process each cross-shard data exchange via the consensus of all nodes in the related shards (i.e., on-chain) to resist a Byzantine environment of blockchain, which eliminates sharding benefits. To tackle the challenge, this paper presents GriDB, the first scalable blockchain database, by designing a novel off-chain cross-shard mechanism for efficient cross-shard database services. Borrowing the idea of off-chain payments, GriDB delegates massive cross-shard data exchange to a few nodes, each of which is randomly picked from a different shard. Considering the Byzantine environment, the untrusted delegates cooperate to generate succinct proof for cross-shard data exchanges, while the consensus is only responsible for the low-cost proof verification. However, different from payments, the database services' verification has more requirements (e.g., completeness, correctness, freshness, and availability); thus, we introduce several new authenticated data structures (ADS). Particularly, we utilize consensus to extend the threat model and reduce the complexity of traditional accumulator-based ADS for verifiable cross-shard queries with a rich set of relational operators. Moreover, we study the necessity of inter-shard load balancing for a scalable blockchain database and design an off-chain and live approach for both efficiency and availability during balancing. An evaluation of our prototype shows the performance of GriDB in terms of scalability in workloads with queries and updates.

The ability to decentralize knowledge graphs (KG) is important to exploit the full potential of the Semantic Web and realize the Web 3.0 vision. However, decentralization also renders KGs more prone to attacks with adverse effects on data integrity and query verifiability. While existing studies focus on ensuring data integrity, how to ensure query verifiability - thus guarding against incorrect, incomplete, or outdated query results - remains unsolved. We propose VeriDKG, the first SPARQL query engine for decentralized knowledge graphs (DKG) that offers both data integrity and query verifiability guarantees. The core of VeriDKG is the RGB-Trie, a new blockchain-maintained authenticated data structure (ADS) facilitating correctness proofs for SPARQL query results. VeriDKG enables verifiability of subqueries by gathering global index information on subgraphs using the RGB-Trie, which is implemented as a new variant of the Merkle prefix tree with an RGB color model. To enable verifiability of the final query result, the RGB-Trie is integrated with a cryptographic accumulator to support verifiable aggregation operations. A rigorous analysis of query verifiability in VeriDKG is presented, along with evidence from an extensive experimental study demonstrating its state-of-the-art query performance on the largeRDFbench benchmark.

Federated Distillation (FD) extends classic Federated Learning (FL) to a more general training framework that enables model-heterogeneous collaborative learning by Knowledge Distillation (KD) across multiple clients and the server. However, existing KD-based algorithms usually require a set of shared input samples for each client to produce soft-prediction for distillation. Worse still, such a manual selection is accompanied by careful deliberations or prior information on clients’ private data distribution, which is not in line with the privacy-preserving characteristic of classic FL. In this paper, we propose a novel training framework to achieve data-independent knowledge transfer by properly designing a distributed generative adversarial network (GAN) between the server and clients that can synthesize shared feature representations to facilitate the FD training. Specifically, we deploy a generator on the server and reuse each local model as a federated discriminator to form a lightweight efficient distributed GAN that can automatically synthesize simulated global feature representations for distillation. Moreover, since the synthesized feature representations are usually more faithful and homologous with global data distribution, faster and better training convergence can be obtained. Extensive experiments on different tasks and heterogeneous models demonstrate the effectiveness of the proposed framework on model accuracy and communication overhead.

As a rising star of social apps, short video apps, e.g., TikTok, have attracted a large number of mobile users by providing fresh and short video contents that highly match their watching preferences. Meanwhile, the booming growth of short video apps imposes new technical challenges on the existing computation and communication infrastructure. Traditional solutions maintain all videos on the cloud and stream them to users via contend delivery networks or the Internet. However, they incur huge network traffic and long delay that seriously affects users’ watching experiences. In this article, we propose an edge-assisted short video sharing framework to address these challenges by caching some highly preferred videos at edge servers that can be accessed by users via high-speed network connections. Since edge servers have limited computation and storage resources, we design an online algorithm with provable approximation ratio to decide which videos should be cached at edge servers, without the knowledge of future network quality and watching preferences changes. Furthermore, we improve the performance by jointly considering video fetching and user-edge association. Extensive simulations are conducted to evaluate the proposed algorithms under various system settings, and the results show that our proposals outperform existing schemes.

The past few years have witnessed an exponential growth of diverse Internet of Things (IoT) devices as well as compelling applications ranging from industrial production to medical care. Dramatic advances in IoT technology not only brought enormous economic opportunities but also challenges (e.g., privacy and security vulnerabilities). Recently, with the appearance of blockchain technology, the integration of IoT and blockchain (BCoT) is considered a promising solution to address these issues. Blockchain provides a secure and scalable data management framework for IoT devices. However, the huge computation and energy cost of the consensus process in blockchain prevents it from being directly applied as a generic platform. To overcome this challenge, in this article, we propose a cloud mining pool-aided BCoT architecture, where the IoT devices can rent the computing resources from the cloud mining pools to offload the mining process. Based on this architecture, we study the mining pool selection problem and analyze the colony behaviors of IoT devices with different pooling strategies. We propose a centralized evolutionary game-based pool selection algorithm for the sake of maximizing the system utility. Considering the non-cooperative relationship among multiple miners, we also propose a lightweight distributed reinforcement learning algorithm, named the ‘WoLF-PHC’ algorithm.

Driven by the burgeoning growth of the Internet of Everything and the substantial breakthroughs in deep learning (DL) algorithms, a booming of artificial intelligence (AI) applications keep emerging. Meanwhile, the advance in existing computing paradigms, i.e., cloud computing and edge computing, provide assorted computing solutions to satisfy the increasingly high requirements for ubiquitous AI services. Nevertheless, there are some non-trivial issues in the computing frameworks, including the underutilization of computing power, the self-interest of computing-power trading mechanism, and the inefficiency of AI services management. To tackle the above issues, we propose a computing-power trading framework based on blockchain, also named AI-Bazaar. In AI-Bazaar, the AI consumers play multiple roles and feel free to contribute the computing power rented from the computing-power provider (CPP) for blockchain mining and AI services. Accordingly, we formulate the computing trading problem as a Stackelberg game. Based on the win or learn fast principle (WoLF), we design a profit-balanced multi-agent reinforcement learning (PB-MARL) algorithm to search the AI-Bazaar equilibrium, while finding the balanced profits for AI consumers and CPP. Numerical simulations are carried out to demonstrate the satisfactory performance and effectiveness of the proposed framework.

As the underlying consensus protocol of Bitcoin and Ethereum blockchains, Proof-of-Work (PoW) features a cryptographic mathematical puzzle whose solution is easy to verify but extremely hard to solve. Under PoW, miners maintain the security of blockchain by devoting computing powers to solve the puzzle; the miner who has solved the puzzle successfully generates a block, along with a reward (e.g., a set of cryptocurrency). The average waiting time to generate a block is inversely proportional to the computing power of the miner. To reduce the average block generation time, a group of individual miners can form a centralized mining pool to aggregate their computing power to solve the puzzle together and share the reward contained in the block. However, if the aggregated computing power of the pool forms a substantial portion of the total computing power in the network, the pooled mining undermines the core spirit of blockchain, i.e., the decentralization, and harms its security. To discourage the pooled mining, we develop a new consensus protocol called Proof-of-Age (PoA) that builds upon the native PoW protocol. The core idea of PoA lies in using Age-of-Work (AoW) to measure the effective mining periods that the miners have devoted to maintaining the security of blockchain. Unlike in the native PoW protocol, in our PoA protocol, miners benefit from its effective mining periods even if they have not successfully mined a block. We first employ a continuous time Markov chain (CTMC) to model the block generation process of the PoA based blockchain. Based on this CTMC model, we then analyze the block generation rates of the mining pool and solo miners respectively. Our analytical results verify that under PoA, the block generation rates of miners in the mining pool are reduced compared to that of solo miners, thereby disincentivizing the pooled mining. Finally, we simulate the mining process in the PoA blockchain to demonstrate the consistency of the analytical results.

The software-defined network (SDN) enabled mobile edge network greatly facilitates network resource management and promotes many emerging applications. However, user mobility may cause the SDN controller to set flow rules frequently, introduce additional flow setup latency, cause delay jitter, and undermine latency-sensitive services. Proactive flow setup is an effective way to eliminate flow setup latency, but existing work fails to maximize the flow setup hit ratio, a metric for evaluating the quality of proactive flow setup decisions, which is critical for latency-sensitive services. In this paper, we study how to proactively set flow rules to maximize the flow setup hit ratio under limited available network resources to eliminate the flow setup latency as much as possible. Then, we formalize the proactive flow setup problem as two integer linear programming problems under two typical routing strategies, default routing and dynamic routing. Both problems are proved to be NP-hard. To tackle these two problems, we propose a linear programming-based polynomial-time approximation algorithm for the default routing case and a greedy-based heuristic algorithm for the dynamic routing case. Extensive trace-driven experimental and simulation results verify that our algorithms can improve the flow setup hit ratio by up to 30.99% compared to existing solutions.

In many applications of the analysis of dynamic graph, manyTiming iterative Graph Processing(TGP) jobs usually need to be generated for the processing of the corresponding snapshots of the dynamic graph to obtain the results at different points of time. For high throughput of such applications, it is expected to run the TGP jobs on the GPU concurrently. Although many GPU-based systems have been recently developed, for out-of-GPU-memory dynamic graph processing, this concurrent way suffers from significant data access overhead due to a large volume of data transfer between CPU and GPU and the interference between these concurrently running jobs, which eventually incurs low GPU utilization ratio. In this work, we observed that the TGP jobs have strong temporal and spatial similarity when they access different snapshots for their own processing as most parts of the snapshots are the same and only a few parts are changing with time. It creates ideal opportunities for efficient concurrent execution of the TGP jobs by dramatically reducing CPU-GPU graph data transfer cost. Based on this observation, we develop the first GPU-based dynamic graph processing systemEGraph, which can be integrated into the existing out-of-GPU-memory static graph processing systems to enable them to efficiently support concurrent execution of TGP jobs on dynamic graphs with the help of GPU accelerators. Different from the existing approaches, we propose in EGraph an effectiveLoading-Processing-Switching(LPS) execution model. It is able to effectively reduce the overhead of CPU-GPU data transfer and ensures a higher GPU utilization ratio for efficient execution of the TGP jobs by fully utilizing the data access similarity between the TGP jobs. Experimental results show that the existing GPU-accelerated systems achieve performance improvements of 2.3-3.5 times after being integrated with EGraph.

Blockchain has been regarded as a trusted carrier for distributed data storage. With large volumes of valuable data stored on blockchain, data query has become a major requirement. However, the existing blockchains do not provide efficient query functionality because of their deep-rooted chain structure. Blockchain database is a new direction that constructs index on top of blockchain to provide rich query functionalities. The existing works are either insecure because the query process separates from the blockchain consensus, or inscalable because all the data needs to be stored in the block. In this paper, we propose a novel semantic blockchain database called MSTDB. We design a hybrid on/off chain blockchain storage architecture in which the majority of blockchain storage is offloaded to the off-chain storage and a novel index structure named Merkle Semantic Trie (MST) is designed to be a secure and semantic bridge between on- and off-chain. Based on MST, MSTDB provides a variety of semantic query functions including multi-keyword query, range query, Top-K query, and cross-chain query. To improve the performance further, we design some index compression and query preprocessing techniques for MSTDB. Extensive experiments demonstrate the effectiveness and efficiency of our blockchain database.

We consider the multiuser detection (MUD) problem, i.e., how to separate and decode colliding data streams, in the uplink of massive Machine Type Communications (mMTC) at millimeter wave (mmWave). Operating on factor-graphs by passing messages, the sum-product algorithm and its variants are widely applied in many other scenarios. However, in this paper, we find that their performance in mMTC at mmWave could be dramatically degraded due to the ill-conditioned MUD channel gain matrix and the existence of enormous short cycles in their corresponding factor-graphs, which are caused by the limited scattering of mmWave and the sharing of a same codebook for error correction among densely located user equipments. Assuming LDPC codes are used for error correction, we further propose a novel sum-product based approach to dealing with the MUD problem in mMTC at mmWave. It first leverages the propagation characteristics of mmWave to optimize the factor-graph for MUD by removing short cycles based on node-split and node-contraction, and then takes a dynamic-programming based method to approximate the messages passing on the resulted factor-graph, which can achieve a higher decoding accuracy. Extensive simulation results show that our approach outperforms the state-of-the-art sum-product based approaches significantly.

Mobile Edge Computing (MEC) has emerged as a promising paradigm catering to overwhelming explosions of mobile applications, by offloading compute-intensive tasks to MEC networks for processing. The surging of deep learning brings new vigor and vitality to shape the prospect of intelligent Internet of Things (IoT), and edge intelligence arises to provision real-time deep neural network (DNN) inference services for users. To accelerate the processing of the DNN inference of a user request in an MEC network, the DNN inference model usually can be partitioned into two connected parts: one part is processed in the local IoT device of the request, and another part is processed in a cloudlet (edge server) in the MEC network. Also, the DNN inference can be further accelerated by allocating multiple threads of the cloudlet to which the request is assigned. In this paper, we study a novel delay-aware DNN inference throughput maximization problem with the aim to maximize the number of delay-aware DNN service requests admitted, by accelerating each DNN inference through jointly exploring DNN partitioning and multi-thread execution parallelism. Specifically, we consider the problem under both offline and online request arrival settings: a set of DNN inference requests is given in advance, and a sequence of DNN inference requests arrives one by one without the knowledge of future arrivals, respectively. We first show that the defined problems are NP-hard. We then devise a novel constant approximation algorithm for the problem under the offline setting. We also propose an online algorithm with a provable competitive ratio for the problem under the online setting. We finally evaluate the performance of the proposed algorithms through experimental simulations. Experimental results demonstrate that the proposed algorithms are promising

Mobile Edge Computing (MEC) promises to provide mobile users with delay-sensitive services at the edge of network, and each user service request usually is associated with a Service Function Chain (SFC) requirement that consists of Virtualized Network Functions (VNFs) in order. The satisfaction of a user on his requested service is heavily impacted by the service reliability. In this paper, we study user satisfaction on services provided by an MEC network through introducing a submodular function based metric to measure user satisfaction. We first formulate a novel user satisfaction problem with the aim to maximize the accumulative user satisfaction, assuming that all available computing resource in the MEC network can be used for service reliability enhancement. We show that the problem is NP-hard, and devise an approximation algorithm with a provable approximation ratio for it. We then consider the problem under a given computing resource budget constraint, for which we devise an approximation algorithm with a provable approximation ratio, at the expense of moderate budget violations. We finally evaluate the performance of the proposed algorithms through experimental simulations. Simulation results demonstrate that the proposed algorithms outperform the comparison baseline algorithms, improving the performance by more$16.1\%$in comparison with the baseline algorithms.

How to efficiently coordinate multiple readers to work together is critical for high throughput in RFID systems. Existing researchs focus on designing efficient reader scheduling strategies that arrange adjacent readers to work in different time to avoid signal collisions. However, the impact of unbalanced tag number of readers on tag read throughput is still challenging. In RFID systems, the distribution of tags is usually variable and uneven, making the number of tags covered by each reader (i.e., the load) imbalanced. This imbalance leads to different execution time for readers: the heavily loaded readers take longer time to collect all tags, while the other readers whose finish execution earlier have to wait in vain. To avoid this useless waiting and improve the system throughput, this paper focuses on the load balancing problem of multiple readers, which is an NP-hard problem. In this paper, we design heuristic algorithms to adjust readers interrogation regions and efficiently balance their loads. The amazing advantage of our algorithm is that it can be adopted by almost all existing protocols in multi-reader systems, including the reader scheduling protocol, to improve system throughput. Extensive experiments demonstrate that our algorithm can significantly improve the throughput in various scenarios.

RFID systems have been widely deployed for various applications such as inventory control, retail management, and object tracking. In many scenarios, coarse-grained system management of tag categories is more common than traditional system management of individual tags. Existing works on tag category management require knowing category IDs in advance, however, few protocols have been specially designed to obtain category IDs in multi-category RFID systems. In this paper, we proposeCIP, a time-efficient protocol specially designed to collect category IDs in multi-category RFID systems. The challenge in solving this problem is to avoid repeatedly collecting category IDs from multiple tags belonging to the same category. We address this challenge by squeezing tags in the same category into the same time slot and letting them reply to the reader simultaneously. Considering that the reader may receive signals from different tags in a slot, we exploit the Manchester coding to decode the aggregated signals. With this design, CIP collects each category ID only once and thus achieves high time efficiency by avoiding redundant identification of the same category ID. We further enhance CIP by utilizing the lightweight vectors to specify the order of tags’ replying and avoid useless slots. Rigorous theoretical analysis is presented to optimize the performance of CIP and guidelines on how to set optimal parameters to minimize the overall execution time are given. We conduct extensive experiments to evaluate the performance of our protocols. The experimental results show that our protocols can significantly improve time efficiency by 67 percent when compared with the state-of-the-art solutions.

Driven by the increasing demand of real-time mobile application processing, Multi-access Edge Computing (MEC) has been envisioned as a promising paradigm for pushing computational resources to network edges. In this paper, we investigate an MEC network enabled by Unmanned Aerial Vehicles (UAV), and consider both the multi-user computation offloading and edge server deployment to minimize the system-wide computation cost under dynamic environment, where users generate tasks according to time-varying probabilities. We decompose the minimization problem by formulating two stochastic games for multi-user computation offloading and edge server deployment respectively, and prove that each formulated stochastic game has at least one Nash Equilibrium (NE). Two learning algorithms are proposed to reach the NEs with polynomial-time computational complexities. We further incorporate these two algorithms into a chess-like asynchronous updating algorithm to solve the system-wide computation cost minimization problem. Finally, performance evaluations based on real-world data are conducted and analyzed, corroborating that the proposed algorithms can achieve efficient computation offloading coupled with proper server deployment under dynamic environment for multiple users and MEC servers.

In this work, we study the problem ofRobustServerPlacement (RSP) for edge computing, i.e., in the presence of uncertain edge server failures, how to determine a server placement strategy to maximize the expected overall workload that can be served by edge servers. We mathematically formulate the RSP problem in the form of robust max-min optimization, derived from two consequentially equivalent transformations of the problem that does not consider robustness and followed by a robust conversion. RSP is challenging to solve, because the explicit expression of the objective function in RSP is hard to obtain, and it is a robust max-min problem with knapsack constraints, which is still an unexplored problem in the literature. We reveal that the objective function is monotone submodular, and propose two solutions to RSP. First, after proving that the involved constraints form a$p$-independence system constraint, where$p$is a parameter determined by the coefficients in the knapsack constraints, we propose an algorithm that achieves a provable approximation ratio in polynomial time. Second, we prove that one of the knapsack constraints is a matroid contraint, and propose another polynomial time algorithm with a better approximation ratio. Furthermore, we discuss the applicability of the aforementioned algorithms to the case with an additional server number constraint. Both synthetic and trace-driven simulation results show that, given any maximum number of server failures, our proposed algorithms outperform four state-of-the-art algorithms and approaches the optimal solution, which applies exhaustive exponential searches, while the proposed latter algorithm brings extra performance gains compared with the former one.

In this paper, a novel task offloading architecture called Flex-MEC is proposed, which achieves efficient task allocation and scheduling (TAS) between MEC servers. By adding metadata before task data, we redesign the offloading process in Flex-MEC, the TAS planning can be conducted without finishing the task data receiving. Once planning is done the task data can be directly forwarded to the allocated server and executed. This reduces latency compared to the traditional way of transmitting, planning, forwarding and executing sequentially. For TAS planning, a multi-server multi-task allocation and scheduling (MMAS) problem is formulated to maximize the MEC system profit. The MMAS problem is proven as an NP-complete problem, thus is challenging to solve. Then, a distributed scheme and a centralized scheme are proposed to solve the MMAS problem with low complexity. In the distributed scheme, the MMAS problem is converted into a non-cooperative game and the existence of Nash Equilibrium (NE) is proven and a low complexity response update algorithm is proposed to converge to NE. And the centralized scheme is based on a greedy idea and runs on a MEC controller in a centralized way. Verified by experiments, these two schemes can achieve better performance than compared schemes.

Blockchain has been leveraged to secure transactions for the m-commerce. However, the intensive computation in the mining process restricts the participation of mobile devices. Currently, some studies have deployed edge computing services to support the mining process, where edge servers managed by one Service Provider (SP) are considered. This paper investigates a more practical scenario with multiple SPs, where servers managed by different SPs have distinct capacities and prices, making miners’ offloading decisions rather complicated. To tackle the above challenges, we consider task offloading, block propagation and miner mobility comprehensively to maximize utilities of miners. Specifically, we first formulate a Markov game, and then design a learning-based offloading algorithm for off-chain computation, where a novel learning model is constructed by integrating Deep Reinforcement Learning (DRL) and Mean Field Theory (MFT) to guarantee a Nash equilibrium. Different from existing studies, each miner merely needs to respond to the average effect from others in our system, insteading of knowing policies of others. Finally, both theoritical and performance results show that our designed algorithm has superiority on average miner utilities and algorithm convergence time compared with other representative algorithms.

Unmanned Aerial Vehicles (UAVs) have been utilized to serve on-ground users with various services, e.g., computing, communication and caching, due to their mobility and flexibility. The main focus of many recent studies on UAVs is to deploy a set of homogeneous UAVs with identical capabilities controlled by one UAV owner/company to provide services. However, little attention has been paid to the issue of how to enable different UAV owners to provide services with differentiated service capabilities in a shared area. To address this issue, we propose a multi-agent imitation learning enabled UAV deployment approach to maximize both profits of UAV owners and utilities of on-ground users. Specially, a Markov game is formulated among UAV owners and we prove that a Nash equilibrium exists based on the full knowledge of the system. For online scheduling with incomplete information, we design agent policies by imitating the behaviors of corresponding experts. A novel neural network model, integrating convolutional neural networks, generative adversarial networks and a gradient-based policy, can be trained and executed in a fully decentralized manner with a guaranteed$\epsilon$-Nash equilibrium. Performance results show that our algorithm has significant superiority in terms of average profits, utilities and execution time compared with other representative algorithms.

Major manufacturers and retailers are increasingly using RFID systems in supply-chain scenarios, where theft of goods during transport typically causes significant economic losses for the consumer. This paper studies how to achieve time-efficient and secure integrity authentication problems in RFID systems. We start with a straightforward solution called SecAuth, which uses a secure identity stored on reserved memory to authenticate tags in a secure way. We then propose a time efficient KTAuth protocol, which design a verification chain mechanism to efficiently verify a small set of key tags using limited on-tag memory. We point out that the limitation of KTAuth is that it takes too much overhead to write a large block of data to tag memory, which leads to the proposed group selection mechanism. The KTAuth with group selection (KTAuth-GS) enables you to select key tags with a single select command, which helps to quickly check the existence of key tags and reduces the data writes on the tags. Experiments and simulation results demonstrate that the proposed KTAuth-GS can defend against counterfeiting attacks by providing more reliable results and reducing the execution time by as much as a factor of 5 when compared with a baseline tag identification protocol.

Lip reading can help people with speech disorders to communicate with others and provide them with a new channel to interact with the world. In this paper, we design and implementHearMe, an accurate and real-time lip-reading system built on commercial RFID devices. HearMe can be used to accurately recognize different words in a pre-defined vocabulary without limitations in light conditions and can be used in multiple user scenarios by leveraging RFID's ability in identifying different users. We design an effective data collection strategy to well capture the tiny and complex signal patterns caused by mouth motion and propose a set of algorithms to extract signal profiles related to mouth motions and mitigate interference factors like multi-path. A carefully designed set of features, including time-domain statistical features and frequency-domain features, are then extracted from the signal to lift the recognition accuracy at the word level. To reduce training costs when the model is used in a new environment, a transfer-learning-based approach is adopted to enhance the robustness of the model in cross-environment scenarios. Experimental results show that HearMe detects speaking actions of the user with an accuracy higher than 0.95 and recognizes different words in a 20-words vocabulary with an average accuracy higher than 0.88. Moreover, the latency of HearMe ($\sim$150ms) is nearly two orders of magnitude less than traditional approaches, making it applicable to practical scenarios that require real-time lip reading.

Gesture recognition based on radio frequency identification (RFID) has attracted much research attention in recent years. Most existing RFID-based gesture recognition approaches use signal profile matching to distinguish different gestures, which incur large recognition latency and fail to support real-time applications. In this paper, we design and implement ReActor, a real-time and accurate gesture recognition system that recognizes a user's gestures with low latency and high accuracy even when the gestures'speed varies. ReActor combines the time-domain statistical features and the frequency-domain features to precisely represent the signal profile corresponding to different gestures. To maintain high accuracy across different environments, we preprocess the signals to remove reflection signals from surrounding objects and use only the signals related to gestures to train the classifier. Moreover, we train a classifier to predict the speed of the gesture and feed the extracted features to different classifiers according to the speed. We implement ReActor and evaluate its performance in different scenarios. Experimental results show that ReActor achieves an average accuracy of 97.2% in recognizing 18 different gestures with an average latency of 72 ms, more than two orders of magnitude faster than approaches based on profile template matching.

Multi-label zero-shot learning extends conventional single-label zero-shot learning to a more realistic scenario that aims at recognizing multiple unseen labels of classes for each input sample. Existing works usually exploit attention mechanism to generate the correlation among different labels. However, most of them are usually biased on severalmajor classeswhile neglect most of theminor classeswith the same importance in input samples, and may thus result in overly diffused attention maps that cannot sufficiently coverminor classes. We argue that disregarding the connection between major and minor classes, i.e., correspond to the global and local information, respectively, is the cause of the problem. In this paper, we propose a novel framework of unbiased multi-label zero-shot learning, by considering various class-specific regions to calibrate the training process of the classifier. Specifically,Pyramid Feature Attention(PFA) is proposed to build the correlation between global and local information of samples to balance the presence of each class. Meanwhile, for the generated semantic representations of input samples, we proposeSemantic Attention(SA) to strengthen the element-wise correlation among these vectors, which can encourage the coordinated representation of them. Extensive experiments on the large-scale multi-label benchmarksMS-COCO,NUS-WIDEandOpen-Imagesdemonstrate that the proposed method surpasses other representative methods by significant margins.

We study the challenging task of malware recognition on both known and novel unknown malware families, called malware open-set recognition (MOSR). Previous works usually assume the malware families are known to the classifier in a close-set scenario, i.e., testing families are the subset or at most identical to training families. However, novel unknown malware families frequently emerge in real-world applications, and as such, require recognizing malware instances in an open-set scenario, i.e., some unknown families are also included in the test set, which has been rarely and nonthoroughly investigated in the cyber-security domain. One practical solution for MOSR may consider jointly classifying known and detecting unknown malware families by a single classifier (e.g., neural network) from the variance of the predicted probability distribution on known families. However, conventional well-trained classifiers usually tend to obtain overly high recognition probabilities in the outputs, especially when the instance feature distributions are similar to each other, e.g., unknown versus known malware families, and thus, dramatically degrade the recognition on novel unknown malware families. To address the problem and construct an applicable MOSR system, we propose a novel model that can conservatively synthesize malware instances to mimic unknown malware families and support a more robust training of the classifier. More specifically, we build upon the generative adversarial networks to explore and obtain marginal malware instances that are close to known families while falling into mimical unknown ones to guide the classifier to lower and flatten the recognition probabilities of unknown families and relatively raise that of known ones to rectify the performance of classification and detection. A cooperative training scheme involving the classification, synthesizing and rectification are further constructed to facilitate the training and jointly improve the model performance. Moreover, we also build a new large-scale malware dataset, named MAL-100, to fill the gap of lacking a large open-set malware benchmark dataset. Experimental results on two widely used malware datasets and our MAL-100 demonstrate the effectiveness of our model compared with other representative methods.

Distributed machine learning (ML) was originally introduced to solve a complex ML problem in a parallel way for more efficient usage of computation resources. In recent years, such learning has been extended to satisfy other objectives, namely, performing learning in situ on the training data at multiple locations and keeping the training datasets private while still allowing sharing of the model. However, these objectives have led to considerable research on the vulnerabilities of distributed learning both in terms of privacy concerns of the training data and the robustness of the learned overall model due to bad or maliciously crafted training data. This article provides a comprehensive survey of various privacy, security, and robustness issues in distributed ML.

In a sharded blockchain, transactions are processed by a number of parallel committees. Thus, the transaction throughput can be largely boosted. A problem is that some groups of blockchain nodes consume large latency to form committees at the beginning of each epoch. Moreover, the heterogeneous processing capabilities of different committees also result in imbalanced consensus latency. Such imbalanced two-phase latency brings a large cumulative age to the transactions pending in transaction pool. Consequently, the blockchain throughput can be significantly degraded. We believe that a good committee-scheduling strategy can reduce the cumulative age of transactions, and thus benefit the throughput. However, we have not yet found a committee-scheduling mechanism that works for accelerating block formation in the context of blockchain sharding. To this end, this paper studies a fine-balanced tradeoff between the transactions’ throughput and their cumulative age in a large-scale sharded blockchain. We formulate this tradeoff as a utility-maximization problem, which is proved NP-hard. To solve this problem, we propose an online distributed Stochastic-Exploration (SE) algorithm, which guarantees a near-optimal system utility. We then rigorously analyze three theoretical properties of the proposed algorithm, including the theoretical convergence time, the probability of committees’ failure due to Sybil attacks, as well as the performance perturbation brought by committees’ offline events. Finally, we evaluate the proposed algorithm using the dataset of real-world blockchain transactions. The simulation results demonstrate that the proposed SE algorithm outperforms other baselines in terms of system utility, the valuable degree of yielded solutions, latency, and throughput performance.

With the rapid development of the Industrial Internet of Things (IIoT), massive IIoT devices connect to industrial networks via wired and wireless. Furthermore, industrial networks pose new requirements on communications, such as strict latency boundaries, ultra-reliable transmission, and so on. To this end, time-sensitive networking (TSN) embedded fifth-generation (5G) wireless communication technology (i.e., TSN-5G networks), is considered the most promising solution to address these challenges. TSN can provide deterministic end-to-end latency and reliability for real-time applications in wired networks. 5G supports ultra-reliable and low-latency communications (uRLLC), providing increased flexibility and inherent mobility support in the wireless network. Thus, the integration of TSN and 5G provides numerous benefits, including increased flexibility, lower commissioning costs, and seamless interoperability of various devices, regardless of whether they use a wired or wireless interface. Nonetheless, the potential barriers between the TSN and 5G systems, such as clock synchronization and end-to-end traffic scheduling, are inevitable. Time synchronization has been studied in many works, so this paper focuses on the end-to-end traffic scheduling problem in TSN-5G networks. We propose a novel integrated TSN and 5G industrial network architecture, where the 5G system acts as a logical TSN-capable bridge. Based on this network architecture, we design a Double Q-learning based hierarchical particle swarm optimization algorithm (DQHPSO) to search for the optimal scheduling solution. The DQHPSO algorithm adopts a level-based population structure and introduces Double Q-learning to adjust the number of levels in the population, which evades the local optimum to further improve the search efficiency. Extensive simulations demonstrate that the DQHPSO algorithm can increase the scheduling success ratio of time-triggered flows compared to other algorithms.

Network slicing is rapidly prevailing in the edge network, which provides computing, network, and storage resources for various services. When the multiple service providers (SPs) respond to their tenants in parallel, individual decisions on the dynamic and shared edge network may lead to resource conflicts, which affects the delivery of network slicing services. Existing works ignore resource interaction and coordination in the multi-SP scenario, which is not in line with the actual situation. Indeed, the complexity of resource interaction caused by the coexistence of multiple SP policies increases the difficulty to solve the formulated optimization model. In this article, we focus on the multi-SP network slicing deployment in parallel. The coordination of network resources between SPs is designed as an effective multi-agent communication mechanism that is merged into multi-agent deep reinforcement learning (MADRL). To deal with dynamic edge networks, we design the neurons hotplugging learning which realizes scalability without a high cost of model retraining. Experiments on real and random networks demonstrate that the proposed multi-SP network slicing mechanism can successfully learn coordination policies and easily adapt to various network scales. It improves the accepted requests by 7.4%, reduces resource conflicts by 14.5%, and shortens the model convergence time by 83.3%.

The surging of deep learning brings new vigor and vitality to shape the prospect of intelligent Internet of Things (IoT), and the rise of edge intelligence enables provisioning real-time deep neural network (DNN) inference services for mobile users. To perform efficient and effective DNN model training in edge computing environments while preserving training data security and privacy of IoT devices, federated learning has been envisioned as an ideal learning paradigm for this purpose. In this article, we study energy-aware DNN model training in edge computing. We first formulate a novel energy-aware, Device-to-Device (D2D) assisted federated learning problem with the aim to minimize the global loss of a training DNN model, subject to bandwidth capacity on an edge server and energy capacity on each IoT device. We then devise a near-optimal learning algorithm for the problem when the training data follows the i.i.d. data distribution. The crux of the proposed algorithm is to explore using the energy of neighboring devices of each device for its local model uploading, by reducing the problem to a series of weighted maximum matching problems in corresponding auxiliary graphs. We also consider the problem without the assumption of the i.i.d. data distribution, for which we propose an efficient heuristic algorithm. We finally evaluate the performance of the proposed algorithms through experimental simulations. Experimental results show that the proposed algorithms are promising.

In the setting of federated optimization, where a global model is aggregated periodically, step asynchronism occurs when participants conduct model training by efficiently utilizing their computational resources. It is well acknowledged that step asynchronism leads to objective inconsistency under non-i.i.d. data, which degrades the model’s accuracy. To address this issue, we propose a new algorithmFedaGrac, which calibrates the local direction to a predictive global orientation. Taking advantage of the estimated orientation, we guarantee that the aggregated model does not excessively deviate from the global optimum while fully utilizing the local updates of faster nodes. We theoretically prove thatFedaGracholds an improved order of convergence rate than the state-of-the-art approaches and eliminates the negative effect of step asynchronism. Empirical results show that our algorithm accelerates the training and enhances the final accuracy.

LEO satellite networks have become a necessary supplement to terrestrial networks aiming to provide worldwide, ubiquitous connectivity, especially in complicated areas (e.g., mountains, oceans, and disaster areas) where terrestrial network infrastructures are typically sparingly distributed or unavailable. However, the increasing computation-intensive Internet-of-Things (IoT) applications (e.g., real-time remote monitoring, intelligent transportation) require not only efficient and reliable communication but also massive computing capabilities. Constrained by the battery and computing resources, the computing tasks and data of applications have to be transmitted to remote cloud servers. This bandwidth limitation and high transmission delay in LEO networks will reduce the quality-of-service (QoS) of IoT applications. Recently, the combination of LEO networks and edge computing (i.e., Satellite Mobile Edge Computing, SMEC) offers significant opportunities to address these problems. The IoT devices can directly get the computing resources directly from satellites rather than remote servers, thus avoiding long-distance transmission. Considering the resource constraints on satellites, offloading policy plays a crucial role in whole system performance. In this paper, we design a hybrid offloading architecture, which applies a centralized training and distributed execution framework. Also, we propose a multi-agent actor-critic reinforcement learning algorithm, where a centralized “critic” is augmented with the global network state to ease the training procedure of distributed user equipments (UE) by evaluating the benefits of their decisions, while the UEs can adjust their policies according to the critic’s evaluation and choose their own decisions relying on their observations.

As a key enabling technology for 5G, network function virtualization abstracts services into software-based service function chains (SFCs), facilitating mission-critical services with high-reliability requirements. However, it is challenging to cost-effectively provide reliable SFCs in dynamic environments due to delayed rewards caused by future SFC requests, limited infrastructure resources, and heterogeneity in hardware and software reliability. Although deep reinforcement learning (DRL) can effectively capture delayed rewards in dynamic environments, its trial-and-error exploration in a vast solution space with massive infeasible solutions may lead to frequent constraint violations and traps in poor local optima. To address these challenges, we propose a RuleDRL algorithm that combines the capability of DRL to capture delayed rewards and the strength of rule-based schemes to explore high-quality solutions without violating constraints. Specifically, we first formulate the reliable SFC provision problem as an integer nonlinear programming problem, which is proven to be NP-hard. Then, we jointly design DRL and rule-based schemes that are coupled to make the final decision and establish a bounded approximation ratio in general cases. Extensive trace-driven simulations show that RuleDRL can save the total cost by up to 65.67% and improve the SFC acceptance ratio by up to 82%, compared to the state-of-the-art solution.

The integrated terrestrial satellite networks (ITSNs) have been deemed as a promising solution to ubiquitous Internet access anytime and anywhere. In this paper, we investigate the spectrum sharing problem in ITSN, in particular focusing on the downlink transmission of the satellite network, which adopts the framing structure for the satellite users (SUs). We model the interference from both terrestrial downlink and uplink transmissions to SUs according to the beamforming techniques. For the terrestrial downlink transmission, we assume that terrestrial users (TUs) are served cooperatively by multiple small base stations (SBSs) via joint beamforming, while the virtual multiple access channel (VMAC) scheme is adopted for the terrestrial uplink transmission. We propose the optimization framework by jointly considering the terrestrial beamformer design and satellite user scheduling to maximize the sum rate of all users. The optimization problems are decomposed into three sub-problems: satellite user scheduling, terrestrial beamformer design, and time slot allocation, which are solved by deep clustering, second-order cone programming (SOCP) (or fractional programming (FP)), and linear programming, respectively. Then, an alternating iterative algorithm is designed to obtain the optimal solution. Simulation results are provided to demonstrate the effectiveness of the proposed algorithm in multiple cases.

Personalized Federated Learning (pFL) not only can capture the common priors from broad range of distributed data, but also support customized models for heterogeneous clients. Researches over the past few years have applied the weighted aggregation manner to produce personalized models, where the weights are determined by calibrating the distance of the entire model parameters or loss values, and have yet to consider the layer-level impacts to the aggregation process, leading to lagged model convergence and inadequate personalization over non-IID datasets. In this paper, we propose a novel pFL training framework dubbed Layer-wised Personalized Federated learning (pFedLA) that can discern the importance of each layer from different clients, and thus is able to optimize the personalized model aggregation for clients with heterogeneous data. Specifically, we employ a dedicated hyper-network per client on the server side, which is trained to identify the mutual contribution factors at layer granularity. Meanwhile, a parameterized mechanism is introduced to update the layer-wised aggregation weights to progressively exploit the inter-user similarity and realize accurate model personalization. Extensive experiments are conducted over different models and learning tasks, and we show that the proposed methods achieve significantly higher performance than state-of-the-art pFL methods.

Traditional one-bit compressed stochastic gradient descent can not be directly employed in multi-hop all-reduce, a widely adopted distributed training paradigm in network-intensive high-performance computing systems such as public clouds. According to our theoretical findings, due to the cascading compression, the training process has considerable deterioration on the convergence performance. To overcome this limitation, we implement a sign-bit compression-based learning synchronization framework, Marsit. It prevents cascading compression via an elaborate bit-wise operation for unbiased sign aggregation and its specific global compensation mechanism for mitigating compression deviation. The proposed framework retains the same theoretical convergence rate as non-compression mechanisms. Experimental results demonstrate that Marsit reduces up to 35% training time while preserving the same accuracy as training without compression.

Payment channel network (PCN) is a promising off-chain technology for blockchain scalability, but it suffers from poor sustainability in practice. In other words, due to the imbalanced transfer in channels, the balance in one direction of channels gradually becomes exhausted until the PCN is rebalanced via a consensus-based rebalancing protocol, during which the involved channels must be suspended. This paper presents Cycle, the first off-chain protocol for a sustainable PCN. It not only keeps the PCN at a balanced level consistently but also avoids the channel freeze incurred by the rebalancing protocol, leading to minimum failed payments and sustained PCN service, respectively. Cycle achieves these benefits based on a novel idea of asynchronous rebalancing. During the normal off-chain running, the participants share the information about their payments and asynchronously rebalance the PCN following the principle that payments along circular channels can cancel each other out. To guarantee security, the protocol resolves the disputes resulting from network latency or malicious participants by a message mechanism for synchronization and a smart contract for arbitration. Moreover, to address the privacy concern during the information sharing, a truncated Laplace mechanism is designed to achieve differential privacy. Finally, we provide a proof-of-concept implementation in Ethereum, over which a real data-based simulation shows that Cycle satisfies 31% more payments than the state-of-the-art technique.

The Low-Earth-Orbit (LEO) mega-constellation networks, by providing low-latency and high-speed communications, are becoming indispensable infrastructures for the future six-generation (6G) architecture. Consequently, the topology, with thousands of satellites equipped with batteries of limited life, has to be adaptively controlled with high energy efficiency. However, existing work lacks the joint consideration of energy efficiency and load adaptation. In this paper, we first propose the line-of-sight condition to determine the candidate ISL set. Next, we model the energy consumption of the LEO mega-constellation networks. Along this direction, we formulate the Load-Adaptive and Energy-Efficient (LAEE) topology control problem in LEO mega-constellation networks and prove its NP-hardness. Finally, we propose the Amortized Energy based Topology Control (AETC) algorithm to solve the LAEE problem, with good adaptation to the fluctuating load and guarantees connectivities between any two satellites. Extensive simulation results demonstrate that the AETC algorithm outperforms related schemes in terms of energy consumption and results in good topology stability.

The space-air-ground (SAG) integrated networks will play a major role in the sixth generation (6G) mobile networks, which will provide global coverage, full connection and pervasive intelligence services for multiple ground Internet of Things (IoT) devices. Moreover, massive computing tasks can be either performed by local devices, or offloaded to edge servers, such as low orbit satellites, high altitude platforms (HAPs) and remote base stations. Nevertheless, the joint computation and communication resource allocation solutions are becoming challenging due to the large-scale state space, time-varying network scenarios, and limited battery capacity. In this paper, we propose a SAG-integrated three-layer heterogenous network model to maximize the sum-rate of ground IoT devices, which further enhances the deep integration of communication and computation resources. Additionally, we develop a Lyapunov-assisted multi-agent proximal policy optimization algorithm to process the task scheduling, HAP selection, battery harvesting, and CPU cycle frequency optimization. Extensive simulation results corroborate that the proposed method has superior performance gains in terms of the remaining battery capacity, energy consumption, and maximum average sum-rate compared with the state-of-the-art baselines.

Federated learning (FL) is a new distributed learning paradigm, in which the clients cooperate to conduct the global model without exposing local private data. However, existing privacy inference attacks on FL show that adversaries can still reverse the training data from the submitted model updates. Recently, secure aggregation has been proposed and integrated into the FL framework, which effectively guarantees privacy through various cryptographic techniques, unfortunately at the cost of a large amount of communication and computation. In this paper, we propose a highly efficient secure aggregation scheme, Fast-Aggregate, which significantly reduces the communication and computation overhead while ensuring data privacy and robustness against clients' dropout. Firstly, Fast-Aggregate employs a multi-group regular graph for efficient secure aggregation to boost data parallelism. Secondly, we leverage polynomial multi-point evaluation and fast Lagrange interpolation methods to handle clients' dropout as well as reduce computational complexity. Finally, we adopt an additive mask to guarantee clients' privacy. Riding on the capabilities of Fast-Aggregate, we achieve the secure aggregation overhead of O (N log2$N$), as opposed to O (N2) in the state-of-the-art works. Besides, Fast-Aggregate improves training speed without loss of model quality and provides flexibility to deal with client corruption at the same time.

Cloud-based services for federated learning has received widespread attention for its ability to collaboratively train a model without collecting users' local data. Although there are existing methods such as homomorphic encryption and secure multi-party computation to address the privacy issues associated with the model parameter exchanging during aggregation, these methods will inevitably lead to huge communication overheads or slow down the training time. Functional encryption (FE) is considered as a new approach to address privacy-preserving federated learning probelms, but the only known FE solution has severe security issues such as leaking master private key, and is impractical. Thus, in this paper, we propose CryptoFE, a cloud-based privacy-preserving federated learning aggregation scheme based on FE. Compared with the only existing FE solution, CryptoFE is efficient in aggregation phase, especially when a high model precision is required, and provides formal privacy guarantees for users' gradients. The experiments with real-world data demonstrate the efficeint performance of our proposed scheme.

Distributed machine learning (DML) has shown great promise in accelerating model training on multiple GPUs. To increase GPU utilization, a common practice is to let multiple learning jobs share GPU clusters, where the most fundamental and critical challenge is how to efficiently schedule these jobs on GPUs. However, existing works about DML job scheduling are constrained to settings with homogeneous GPUs. GPU heterogeneity is common in practice, but its influence on multiple DML job scheduling has been seldom studied. Moreover, DML jobs have internal structures that contain great parallelism potentials, which have not yet been fully exploited in the heterogeneous computing environment. In this paper, we propose Hare, a DML job scheduler that exploits both inter-job and intra-job parallelism in a heterogeneous GPU cluster. Hare has three novel designs. First, Hare optimizes GPU execution environment to reduce task switching overhead by exploiting unique features of DML scheduling. Second, Hare adopts a relaxed fixed-scale synchronization scheme that allows independent tasks to be flexibly scheduled within a training round. Finally, we propose a fast heuristic algorithm to minimize the total weighted job completion time by jointly considering job features and hardware heterogeneity. Its theoretical bound is derived. We evaluate Hare using a small-scale testbed and a trace-driven simulator. The results show that it can outperform the state-of-the-art by about 2x.

Wireless systems are upgraded to use green energy (e.g., solar, wind, and tide energy) such that the greenhouse gas emission can be neutralized. This work incorporates the on-grid energy into a green coordinated multi-point (CoMP) system to handle the volatile arrival of green energy. In the green CoMP, the long-term weighted throughput maximization problem is investigated by expecting a non-positive consumption of the long-term on-grid energy. Motivated by the capacity-achieving property and simple implementation, an online zero-forcing dirty paper precoder is proposed to update the precoding matrices by combining statistical learning with the Lyapunov learning. A tradeoff relation is theoretically established to show that the long-term weighted throughput approaches the $\mathcal{O}(V)$ -neighbor of optimal value while the long-term consumed on-grid energy increases at a rate of $\mathcal{O}\left( {{{\log }^2}(V)/\sqrt V } \right)$, where V is an introduced control parameter. Numerical results are used to verify the performance of the online zero-forcing dirty paper precoder.

Age of Information (AoI) has emerged as a new metric to measure data freshness from the destination’s perspective. The problem of optimizing AoI has been attracting extensive interests recently. However, existing works mainly focused on scheduling data transmission for AoI optimization. While at wireless-powered network edge, the charging plan of source nodes also requires to be computed in advance, which means the system AoI is determined by not only the data transmission decision but also the charging plan. Thus, in this paper, we investigate the first work to optimize the weighted peak AoI from the point of charging at wireless-powered network edge with a directional charger. Firstly, to minimize the weighted sum of average peak AoI, the AoI minimization problem is transformed to a charging time optimization problem with respect to the overlapped charging areas and average peak AoI, and an approximate algorithm is proposed to obtain the required charging time for each source node. Then, an age-based scheduling algorithm is proposed to compute the charging and data transmission decisions for each source node simultaneously, which can not only optimize the weighted sum of average peak AoI but also guarantee the maximum peak AoI for each source node. The proposed algorithm is proved to have an approximation ratio of up to (1+φ), where φ is a much smaller value related to the weight of each source node. Finally, the simulation results verify the high performance of proposed algorithms in terms of average and maximum peak AoI.

Federated learning (FL) clients may be reluctant to participate in the energy-consuming FL unless they are incentivized. Existing incentive mechanisms seldom consider the economic properties, e.g., social welfare, individual rationality and incentive compatibility, which significantly limits the sustainability of FL to attract more clients. The Vickrey–Clarke–Groves (VCG) auction is an ideal mechanism for simultaneously guaranteeing all crucial economic properties to maximize social welfare. However, VCG auction cannot be applied directly to FL scenarios due to the following challenges: 1) It requires precise analytical derivation of the optimal strategy, which is unavailable due to the inherent model-unknown and privacy-sensitive characteristics of FL. 2) Current auction modeling decomposes the entire process into multiple independent rounds and solves them one-by-one, which breaks the successive correlation between rounds in the long-term training process of FL. To overcome these challenges, this paper presents a long-term online VCG auction mechanism for FL that employs an experience-driven deep reinforcement learning algorithm to obtain the optimal strategy. Besides, we extend long-term forms of the crucial economic properties for the successive FL process. Furthermore, knowledge transfer is applied to reduce the excessive training overhead arising from the VCG payment rules. By exploiting the environmental similarity among sub-auctions, we develop the strategy sharing to significantly cut the training time by half. Finally, we theoretically prove the extended economic properties and conduct extensive experiments on multiple real-world datasets. Compared with state-of-the-art approaches, the long-term social welfare of FL increases by 36% with a 37% reduction in payment.

Network slicing is rapidly prevailing in edge cloud, which provides computing, network and storage resources for various services. When the multiple service providers (SPs) respond to their tenants in parallel, individual decisions on the dynamic and shared edge cloud may lead to resource conflicts. The resource conflicts problem can be formulated as a multi-objective constrained optimization model; however, it is challenging to solve it due to the complexity of resource interactions caused by co-existing multi-SP policies. Therefore, we propose a CommDRL scheme based on multi-agent deep reinforcement learning (MADRL) and multi-agent communication to tackle the challenge. CommDRL can coordinate network resources between SPs with less overhead. Moreover, we design the neurons hotplugging learning in CommDRL to deal with dynamic edge cloud, which realizes scalability without a high cost of model retraining. Experiments demonstrate that CommDRL can successfully obtain deployment policies and easily adapt to various network scales. It improves the accepted requests by 7.4%, reduces resource conflicts by 14.5%, and shortens the model convergence time by 83.3%.

The prevalent personalized federated learning (PFL) usually pursues a trade-off between personalization and generalization by maintaining a shared global model to guide the training process of local models. However, the sole global model may easily transfer deviated context knowledge to some local models when multiple latent contexts exist across the local datasets. In this paper, we propose a novel concept called contextualized generalization (CG) to provide each client with fine-grained context knowledge that can better fit the local data distributions and facilitate faster model convergence, based on which we properly design a framework of PFL, dubbed CGPFL. We conduct detailed theoretical analysis, in which the convergence guarantee is presented and a speedup of order 1/2 w.r.t the number of contexts over most existing methods is granted. To quantitatively study the generalization-personalization trade-off, we introduce the generalization error measure and prove that the proposed CGPFL can achieve a better trade-off than existing solutions. Moreover, our theoretical analysis further inspires a heuristic algorithm to find a near-optimal trade-off in CGPFL. Experimental results on multiple real-world datasets show that our approach surpasses the state-of-the-art methods on test accuracy by a significant margin.

Recent studies have shown that the training samples can be recovered from gradients, which are called Gradient Inversion (GradInv) attacks. However, there remains a lack of extensive surveys covering recent advances and thorough analysis of this issue. In this paper, we present a comprehensive survey on GradInv, aiming to summarize the cutting-edge research and broaden the horizons for different domains. Firstly, we propose a taxonomy of GradInv attacks by characterizing existing attacks into two paradigms: iteration- and recursion-based attacks. In particular, we dig out some critical ingredients from the iteration-based attacks, including data initialization, model training and gradient matching. Second, we summarize emerging defense strategies against GradInv attacks. We find these approaches focus on three perspectives covering data obscuration, model improvement and gradient protection. Finally, we discuss some promising directions and open problems for further research.

State-of-the-art blockchain sharding solutions, say Monoxide, can induce imbalanced transaction (TX) distributions among all blockchain shards due to their account deployment mechanisms. Imbalanced TX distributions then cause hot shards, in which the cross-shard TXs may experience an unlimited length of confirmation latency. Thus, how to address the hot-shard issue and how to reduce cross-shard TXs become significant challenges of blockchain state sharding. Through reviewing the related studies, we find that a cross-shard TX protocol that can achieve workload balance among all shards and simultaneously reduce the number of cross-shard TXs is still absent from the literature. To this end, we propose BrokerChain, which is a cross-shard blockchain protocol devised for the account/balance-based state sharding. Essentially, BrokerChain exploits fine-grained state partition and account segmentation. We also elaborate on how BrokerChain handles cross-shard TXs through broker accounts. The security issues and other properties of BrokerChain are analyzed substantially. Finally, we conduct comprehensive evaluations using both a cloud-based prototype and a transaction-driven simulator. The evaluation results show that BrokerChain outperforms other solutions in terms of system throughput, transaction confirmation latency, the queue size of transaction pool, and workload balance.

Federated Learning (FL) is susceptible to gradient leakage attacks, as recent studies show the feasibility of obtaining private training data on clients from publicly shared gradients. Existing work solves this problem by incorporating a series of privacy protection mechanisms, such as homomorphic encryption and local differential privacy to prevent data leakage. However, these solutions either incur significant communication and computation costs, or significant training accuracy loss. In this paper, we show that the sensitivity of gradient changes w.r.t. training data is an essential measure of information leakage risk. Based on this observation, we present a novel defense, whose intuition is perturbing gradients to match information leakage risk such that the defense overhead is lightweight while privacy protection is adequate. Our another key observation is that global correlations of gradients could compensate for this perturbation. Based on such compensation, training can achieve guaranteed accuracy. We conduct experiments on MNIST, Fashion-MNIST and CIFAR-10 for defending against two gradient leakage attacks. Without sacrificing accuracy, the results demonstrate that our lightweight defense can decrease the PSNR and SSIM between the reconstructed images and raw images by up to more than 60% for both two attacks, compared with baseline defensive methods.

Serverless edge computing is booming as an efficient carrier of deploying complex applications composed of dependent functions, whose assignment decisions highly influence the application performance. Although similar problem has been widely studied, none of existing approaches considers the diversity of communication styles, which is specially introduced in serverless computing and also imposes high influence to the performance efficiency. We compare two communication styles, called direct-passing and remote-storage, to transmit intermediate data between functions. We find that there is no single communication style that can prevail under all scenarios and the optimal selection depends on several factors, such as fanout degree, data size, and network bandwidth. Hence, how to select the appropriate communication style for each inter-function communication link, together with the function assignment decision, is essential to the application performance. To this end, we propose a Priority-based ASsignment and Selection (PASS) algorithm with joint consideration of function assignment and communication style selection. We theoretically analyze the approximation ratio of PASS algorithm and extensive experiments on real-world applications show that PASS can averagely reduce the completion time by 24.1% in comparison with state-of-the-art approaches.

The surging of deep learning brings new vigor and vitality to shape the prospect of intelligent Internet of Things (IoT), and edge intelligence arises to provision real-time deep neural network (DNN) inference services for mobile users. To perform efficient and effective DNN model training in edge environments while preserving training data security and privacy of IoT devices, federated learning has been envisioned as an ideal learning paradigm for this purpose. In this paper we study energy-aware DNN model training in an edge environment. We first formulate a novel energy-aware, device-to-device (D2D) assisted federated learning problem with the aim to minimize the global loss of a training DNN model, subject to bandwidth capacity on an edge server and the energy capacity on each IoT device. We then devise an efficient heuristic algorithm for the problem. The crux of the proposed algorithm is to explore the energy usage of neighboring devices of each device for its local model uploading, by reducing the problem to a series of maximum weight matching problems in corresponding auxiliary graphs. We finally evaluate the performance of the proposed algorithm through experimental simulations. Experimental results show that the proposed algorithm is promising.

It witnesses that the collaborative learning (CL) systems often face the performance bottleneck of limited bandwidth, where multiple low-end devices continuously generate data and transmit intermediate features to the cloud for incremental training. To this end, improving the communication efficiency by reducing traffic size is one of the most crucial issues for realistic deployment. Existing systems mostly compress features at pixel level and ignore the characteristics of feature structure, which could be further exploited for more efficient compression. In this paper, we take new insights into implementing scalable CL systems through a hierarchical compression on features, termed Stripe-wise Group Quantization (SGQ). Different from previous unstructured quantization methods, SGQ captures both channel and spatial similarity in pixels, and simultaneously encodes features in these two levels to gain a much higher compression ratio. In particular, we refactor feature structure based on inter-channel similarity and bound the gradient deviation caused by quantization, in forward and backward passes, respectively. Such a double-stage pipeline makes SGQ hold a sublinear convergence order as the vanilla SGD-based optimization. Extensive experiments show that SGQ achieves a higher traffic reduction ratio by up to 15.97 times and provides 9.22 times image processing speedup over the uniform quantized training, while preserving adequate model accuracy as FP32 does, even using 4-bit quantization. This verifies that SGQ can be applied to a wide spectrum of edge intelligence applications.

Sharding technique is viewed as the most promising solution to improving blockchain scalability. However, to implement a sharded blockchain, developers have to address two major challenges. The first challenge is that the ratio of cross-shard transactions (TXs) across blockchain shards is very high. This issue significantly degrades the throughput of a blockchain. The second challenge is that the workloads across blockchain shards are largely imbalanced. If workloads are imbalanced, some shards have to handle an overwhelming number of TXs and become congested very possibly. Facing these two challenges, a dilemma is that it is difficult to guarantee a low cross-shard TX ratio and maintain the workload balance across all shards, simultaneously. We believe that a fine-grained account-allocation strategy can address this dilemma. To this end, we first formulate the tradeoff between such two metrics as a network-partition problem. We then solve this problem using a community-aware account partition algorithm. Furthermore, we also propose a sharding protocol, named Transformers, to apply the proposed algorithm into the sharded blockchain system. Finally, trace-driven evaluation results demonstrate that the proposed protocol outperforms other baselines in terms of throughput, latency, cross-shard TX ratio, and the queue size of transaction pool.

We explore the problem of selectively forgetting categories from trained CNN classification models in federated learning (FL). Given that the data used for training cannot be accessed globally in FL, our insights probe deep into the internal influence of each channel. Through the visualization of feature maps activated by different channels, we observe that different channels have a varying contribution to different categories in image classification.

With the development of smart cities, the chimney construction method can no longer meet service needs. It is extremely urgent to build a unified urban brain, and the core issue is data sharing and fusion. Aiming at the problems of data island, data leakage, and high trust cost in the IoT of the smart city, a lightweight and trusted sharing mechanism (LTSM) is proposed. First, the blockchain is combined with federated learning to realize the data sharing, which not only protects the private data, but also ensures the sharing process trust. Then, a node selection algorithm based on credit value and a node evaluation algorithm based on smart contract are designed to improve the quality of federated learning. Finally, we propose an improved raft consensus to meet the delay and security requirements of the consortium blockchain in the smart city scenario. In the simulation, we evaluate the federated learning algorithm, the node selection algorithm, and the improved raft consensus, respectively. The experimental results show that the LTSM mechanism has a good application value. The federated learning model has a better accuracy, but its training time is also longer. The node selection algorithm is helpful to improve the accuracy of the federated learning model. The improved raft consensus improves the throughput.

Federated Learning has attracted widespread attention in recent years because it allows massive edge nodes to collaboratively train machine learning models without sharing their private data sets. However, these edge nodes are usually heterogeneous in computational capability and statistically different in data distribution, i.e., non-independent and identically distributed (IID), leading to significant performance degradation. Although status quo asynchronous training methods can solve the heterogeneity issue, they cannot prevent the non-IID problem from reducing the convergence rate. In this article, we propose a novel paradigm that schedules the gradient with partially averaged gradients and applies the global momentum (GSGM) for asynchronous training over non-IID data sets in an edge environment. Our key idea is to apply global momentum and partial average on the biased gradients calculated on edge nodes after scheduling, to make the training process stable. Empirical results demonstrate that GSGM can well adapt to different degrees of non-IID data and bring 20% performance gains in terms of training stability for popular optimization algorithms with enhanced accuracy over Fashion-Mnist and CIFAR-10 data sets.

The ever-growing artificial intelligence (AI) applications have greatly reshaped our world in many areas, e.g., smart home, computer vision, natural language processing, etc. Behind these applications are usually machine learning (ML) models with extremely large size, which require huge data sets for accurate training to mine the value contained in the big data. Large ML models, however, can consume tremendous computing resources to achieve decent performance and thus, it is difficult to train them in resource-constrained Internet of Things (IoT) environments, which would prevent further development and application of AI techniques in the future. To deal with such challenges, there are many efforts on accelerating the training process for large ML models in IoT. In this article, we provide a comprehensive review on the recent advances toward reducing the computing cost during the training stage while maintaining comparable model accuracy. Specifically, the optimization algorithms that aim to improve the convergence rate are emphasized over various distributed learning architectures that exploit ubiquitous computing resources. Then, the article elaborates the computation hardware acceleration and communication optimization for collaborative training among multiple learning entities. Finally, the remaining challenges, future opportunities, and possible directions are discussed.

Path planning is critical to realizing a high-performance unmanned aerial vehicle (UAV) crowdsensing system, which can be deployed to carry out large-scale tasks in the physical world, especially in emergency scenarios, such as earthquakes and mudslides. Deep reinforcement learning (DRL) has recently proven its superiority in path design. However, it is often applied under the assumption that the entire status of the target region is available, which is hard to achieve in practice. Instead, efforts should be made to ensure the efficient flight of several UAVs in order to collect data with incomplete observations in specified places. In this work, we set out to create a high-performance UAV crowdsensing system by combining DRL with partial observations. We present a novel DRL-based path-planning algorithm called DRL-PP. Specifically, we integrate an attention mechanism into the actor–critic technique to assist UAV swarm collaboration to collect data. We also design an incentive mechanism to ease the problem of sparse reward. Furthermore, we provide a dilemma detection system to prevent the generation of overlapping flight paths. Experimental results from extensive simulations prove that compared with the state-of-the-art approaches, the proposed DRL-PP can significantly improve the efficiency of data collection.

Federated learning (FL) is an emerging paradigm through which decentralized devices can collaboratively train a common model. However, a serious concern is the leakage of privacy from exchanged gradient information between clients and the parameter server (PS) in FL. To protect gradient information, clients can adopt differential privacy (DP) to add additional noises and distort original gradients before they are uploaded to the PS. Nevertheless, the model accuracy will be significantly impaired by DP noises, making DP impracticable in real systems. In this work, we propose a novel noise information secretly sharing (NISS) algorithm to alleviate the disturbance of DP noises by sharing negated noises among clients. We theoretically prove that: 1) if clients are trustworthy, DP noises can be perfectly offset on the PS and 2) clients can easily distort negated DP noises to protect themselves in case that other clients are not totally trustworthy, though the cost lowers model accuracy. NISS is particularly applicable for FL across multiple Internet of Things (IoT) systems, in which all IoT devices need to collaboratively train a model. To verify the effectiveness and the superiority of the NISS algorithm, we conduct experiments with the MNIST and CIFAR-10 data sets. The experimental results verify our analysis and demonstrate that NISS can improve model accuracy by 19% on average and obtain better privacy protection if clients are trustworthy.

As a promising solution to blockchain scalability, sharding divides blockchain nodes into small groups called shards, splitting the workload. Existing works for sharding, however, are limited by cross-shard transactions, since they need to split each cross-shard transaction into multiple sub-transactions, each of which costs a consensus round to commit. In this paper, we introduce PYRAMID, a novel sharding system based on the idea of layered sharding. In PYRAMID, the nodes with better hardware are allowed to participate in multiple shards and store the blockchains of these shards thus they can validate and execute the cross-shard transactions without splitting. Next, to commit the cross-shard transactions with consistency among the related shards, we design a cooperative cross-shard consensus based on collective signature-based inter-shard collaboration. Furthermore, we present an optimization framework to compute an optimal layered sharding strategy maximizing the transaction throughput with the constraint of system security and node resource. Finally, we implement a prototype for PYRAMID based on Ethereum and the experimental results reveal the efficiency of PYRAMID in terms of performance and scalability, especially in workloads with a high percentage of cross-shard transactions. PYRAMID improves the throughput by up to 3.2 times compared with the state-of-the-art works and achieves about 3821 transaction per seconds for 20 shards.

Recent decades have seen a significant increase in the frequency, intensity, and impact of natural disasters and other emergencies, forcing the governments around the world to make emergency response and disaster management national priorities. The growth of extremely large and complex datasets—commonly referred to asbig data—and various advances in information and communications technology and computing now support more effective approaches to humanitarian relief, logistical coordination, overall disaster management, and long-term recovery in connection with natural disasters and emergency events. Leveraging big data and technological advances for emergency management has attracted considerable attention in the research community. However, the desired merging ofbig data and emergency management(BDEM) requires coordinated efforts to align and define interdisciplinary terminologies and methodologies. To date, the key concepts and technologies in this emerging research area have not been coherently discussed in a sufficiently broad and multidisciplinary manner. In this article, an international team presents an overview of the BDEM domain, highlighting a general framework and discussing key challenges from several perspectives. We introduce and summarize typical technologies and applications, organized into the six broad categories of remote sensing, resilient communication networks, mobile communication networks, human mobility and urban sensing, social network analysis, and knowledge graphs. Finally, we outline several directions of future research.

Data collection in big data is an effective way to aggregate information that the collector is interested in. However, there is no assurance for the data that the users provide. Since collector does not have the ability to check the authenticity of every piece of information, the trustworthiness of users participated in the collection become important. In this paper, we design an efficient approach to calculate users’ trustworthiness in data collection for big data context. We divide the trustworthiness into familiarity trustworthiness and similarity trustworthiness, and study the influences of user actions on trustworthiness. To prevent malicious users from raising their trustworthiness and providing false information that may mislead final results, we also design a security queue to record users’ historical trust information, so that we can detect malicious users with high accuracy. Simulation results show that our model can sensitively resist the malicious actions of users.

Distributed gradient descent has been widely adopted in the machine learning field because considerable computing resources are available when facing the huge volume of data. Specifically, the gradient over the whole data is cooperatively computed by multiple workers. However, its performance can be severely affected by slow workers, namely stragglers. Recently, coding-based approaches have been introduced to mitigate the straggler problem, but they could hardly deal with the heterogeneity among workers. Besides, they always discard the results of stragglers causing huge resource waste. In this article, we first investigate how to tolerate stragglers by discarding their results and then seek to leverage the stragglers. For tolerating stragglers, we propose a heterogeneity-aware coding scheme that encodes gradients adaptive to the computing capability of workers. Theoretically, this scheme is optimal for stragglers tolerance. Relying on the scheme, we further propose an algorithm called DHeter-aware to exploit the gradients of stragglers which we called delayed gradients. Moreover, theoretical results characterized for DHeter-aware exhibits the same convergence rate as the gradient descent without delayed gradients. Experiments on various tasks and clusters demonstrate that our coding scheme outperforms all the state-of-the-art methods and the DHeter-aware further accelerates the coding scheme by achieving 25 percent time savings.

As the large-scale deployment of machine learning applications, there is much research attention on exploiting a vast amount of data stored on mobile clients. To preserve data privacy, federated learning has been proposed to enable large-scale machine learning by massive clients without exposing raw data. Existing works of federated learning struggle for accelerating the learning process, but ignore the energy efficiency that is critical for resource-constrained clients. In this article, we propose to improve the energy efficiency of federated learning by lowering CPU cycle frequencies of clients who are faster in the training group. Based on this idea, we formulate an optimization problem aiming to minimize the total system cost defined as a weighted sum of learning time and energy consumption. Due to the hardness of the formulated optimization problem and unpredictability of network quality, we propose L4L (Learning for Learning), an experience-driven computational resource control approach based on the deep reinforcement learning, which can derive the near-optimal solution with only the clients’ bandwidth information in the previous training rounds. We conduct the experiments using both real-world traces and synthetic traces to evaluate the proposed L4L approach. The results demonstrate the superiority of L4L as compared with the state-of-the-art solutions.

Federated learning (FL) has been widely recognized as a promising approach by enabling individual end-devices to cooperatively train a global model without exposing their own data. One of the key challenges in FL is the non-independent and identically distributed (Non-IID) data across the clients, which decreases the efficiency of stochastic gradient descent (SGD) based training process. Moreover, clients with different data distributions may cause bias to the global model update, resulting in a degraded model accuracy. To tackle the Non-IID problem in FL, we aim to optimize the local training process and global aggregation simultaneously. For local training, we analyze the effect of hyperparameters (e.g., the batch size, the number of local updates) on the training performance of FL. Guided by the toy example and theoretical analysis, we are motivated to mitigate the negative impacts incurred by Non-IID data via selecting a subset of participants and adaptively adjust their batch size. A deep reinforcement learning based approach has been proposed to adaptively control the training of local models and the phase of global aggregation. Extensive experiments on different datasets show that our method can improve the model accuracy by up to 30 percent, as compared to the state-of-the-art approaches.

Mobile edge computing (MEC) has emerged as a promising technology that offers resource-intensive yet delay-sensitive applications from the edge of mobile networks. With the emergence of complicated and resource-hungry mobile applications, offloading user tasks to cloudlets of nearby mobile edge-cloud networks is becoming an important approach to leverage the processing capability of mobile devices, reduce mobile device energy consumptions, and improve experiences of mobile users. In this article we first study the provisioning of virtualized network function (VNF) services for user requests in an MEC network, where each user request has a demanded data packet rate with a specified network function service requirement, and different user requests need different services that are represented by virtualized network functions instantiated in cloudlets. We aim to maximize the number of user request admissions while minimizing their admission cost, where the request admission cost consists of the computing cost on instantiations of requested VNF instances and the data packet traffic processing of requests in their VNF instances, and the communication cost of routing data packet traffic of requests between users and the cloudlets hosting their requested VNF instances. We study the joint VNF instance deployment and user requests assignment in MEC, by explicitly exploring a non-trivial usage tradeoff between different types of resources. To this end, we first formulate the cost minimization problem that admits all requests by assuming that there is sufficient computing resource in MEC to accommodate the requested VNF instances of all requests, for which we formulate an Integer Linear Programming solution and two efficient heuristic algorithms. We then deal with the problem under the computing resource constraint. We term this problem as the throughput maximization problem by admitting as many as requests, subject to computing resource capacity on each cloudlet, for which we formulate an ILP solution when the problem size is small; otherwise, we devise efficient algorithms for it. We finally evaluate the performance of the proposed algorithms through experimental simulations. Experimental results demonstrate that the proposed algorithms are promising. To the best of our knowledge, we are the first to explicitly explore the usage tradeoff between computing and communication resources in the admissions of user requests in MEC through introducing a novel load factor concept to minimize the request admission cost and maximize the network throughput.

Recently, the term ‘Internet of Things’ (IoT) has garnered great attention. As a trusted, dependable, and decentralized approach, blockchain has already been used in IoT. However, the existing blockchain has a number of drawbacks that prevent it from being used as a generic platform for IoT. The nodes in IoT are heavily resource-limited, especially computing and networking resources. Unfortunately, they are necessary for the blockchain to solve complicated puzzles and propagate blocks. In this paper, we propose agent mining and cloud mining approaches to solve the above problem in the blockchain-enabled IoT. To be specific, miners act as mining agents for nodes in IoT, offload mining tasks to cloud computing servers, and use networking resources dynamically. Furthermore, in order to enhance the performance, the access selection of users, computing resources allocation, and networking resources allocation are formulated as a joint optimization problem. We then propose a dueling deep reinforcement learning approach to address this problem. Numerical results justify the effectiveness of our proposed scheme.

Distributed deep learning has been widely used in training deep neural networks, especially for big models on massive datasets. Parameter Server (PS) architecture is the most popular distributed training framework, which can flexibly design the global parameter update manner. However, when scaling to complex heterogeneous clusters, stragglers make it difficult for existing distributed paradigms on PS framework to balance between synchronous waiting and staleness, which slows down the model training sharply. In this article, we proposeGrouping Stale Synchronous Parallel(GSSP) scheme, which groups workers with similar performance together. Group servers coordinate intra-group workers using Stale Synchronous Parallel while they communicate with each other asynchronously to eliminate stragglers and refine the model weights. We further proposeGrouping Dynamic Tok-K Sparsification(GDTopK), which dynamically adjusts the upload ratio for each group so as to make communication volume differentiated and mitigate inter-group iteration speed gap. We have conducted experiments on LeNet-5 on MNIST, ResNet-18, VGG-19 on Cifar-10, and Seq2Seq on Multi30k. Results show that GSSP accelerates the training by 46%$\sim$120%, with less than 1 percent accuracy drop. And GDTopK can make up for part of the lost accuracy.

The paradigm of Internet of Things (IoT) has attracted much attention in both academia and industry areas in the past decades. Recently, the integration of IoT and social networks is advocated to promote further development. This article focuses on the applications, solutions, and challenges of Social Internet of Things (SIoT) over collaborative edge computing, which exploits the advantages of both mobile edge computing and social relationships among SIoT users. First, applications of collaborative offloading, caching, and streaming data processing are illustrated, followed by representative social-aware solutions, including auction, coalition game, and federated learning. Finally, several research challenges are described. The main contributions of this article are as follows: 1) we specify the role of social ties in traditional applications of IoT and its impacts on individual’s selection; 2) we elaborate the reasons why the presented three approaches can be applied in SIoT; and 3) the discussed challenges can contribute to the future development of secure, robust, and intelligent SIoT frameworks.

Mobile edge computing (MEC) has emerged as a promising technology to push the cloud frontier to the network edge, provisioning network services in proximity of mobile users. Serving users at edge clouds can reduce service latency, lower operational cost, and improve network resource availability. Along with the MEC technology, network function virtualization (NFV) is another promising technique that implements various network service functions as pieces of software in cloudlets (servers or clusters of servers). Providing virtualized network service for mobile users can improve user service experience, simplify network service deployment, and ease network resource management. However, mobile users move in networks arbitrarily, and different users usually request different services with different resource demands and delay requirements. It thus poses a great challenge to providing reliable and seamless virtualized network services for mobile users in an MEC network while meeting their individual delay requirements, subject to resource capacities on the network. In this paper, we focus on the provisioning of virtualized network function services for mobile users in MEC that takes into account user mobility and service delay requirements. We first formulate two novel optimization problems of user service request admissions with the aims to maximize the accumulative network utility and accumulative network throughput for a given time horizon, respectively. We then devise a constant approximation algorithm for the utility maximization problem. We also develop an online algorithm for the accumulative throughput maximization problem. We finally evaluate the performance of the proposed algorithms through experimental simulations. Experimental results demonstrate that the proposed algorithms are promising.

Intelligent Transportation System (ITS) is critical to cope with traffic events, e.g., traffic jams and accidents, and provide services for personal traveling. However, existing researches have not jointly considered the user data safety, utility and system latency comprehensively, to the best of our knowledge. Since both safe and efficient transmissions are significant for ITS, we construct a blockchain-enabled crowdsensing framework for distributed traffic management. First, we illustrate the system model and formulate a multi-objective optimization problem. Due to its complexity, we decompose it into two subproblems, and propose the corresponding schemes, i.e., a Deep Reinforcement Learning (DRL)-based algorithm and a DIstributed Alternating Direction mEthod of Multipliers (DIADEM) algorithm. Extensive experiments are carried out to evaluate the performance of our solutions, and experimental results demonstrate that the DRL-based algorithm can legitimately select active miners and transactions to make a satisfied trade-off between the blockchain safety and latency, and the DIADEM algorithm can effectively select task computation modes for vehicles in a distributed way to maximize their social welfare.

Federated Learning (FL) is a promising machine learning paradigm to cooperatively train a global model with highly distributed data located on mobile devices. Aiming to optimize the communication efficiency for gradient aggregation and model synchronization among large-scale devices, we propose a relay-assisted FL framework. By breaking the traditional transmission-order constraint and exploiting the broadcast characteristic of relay nodes, we design a novel synchronization scheme named Partial Synchronization Parallel (PSP), in which models and gradients are transmitted simultaneously and aggregated at relay nodes, resulting in traffic reduction. We prove that PSP has the same convergence rate as the sequential synchronization approaches via rigorous analysis. To further accelerate the training process, we integrate PSP with any unbiased and error-bounded compression technologies and prove that the convergence properties of the resulting scheme still hold. Extensive experiments are conducted in a distributed cluster environment with real-world datasets and the results demonstrate that our proposed approach reduces the training time up to 37 percent compared to state-of-the-art methods.

Vehicular edge computing (VEC) is a promising paradigm based on the Internet of vehicles to provide computing resources for end users and relieve heavy traffic burden for cellular networks. In this paper, we consider a VEC network with dynamic topologies, unstable connections and unpredictable movements. Vehicles inside can offload computation tasks to available neighboring VEC clusters formed by onboard resources, with the purpose of both minimizing system energy consumption and satisfying task latency constraints. For online task scheduling, existing researches either design heuristic algorithms or leverage machine learning, e.g., deep reinforcement learning (DRL). However, these algorithms are not efficient enough because of their low searching efficiency and slow convergence speeds for large-scale networks. Instead, we propose an imitation learning enabled online task scheduling algorithm with near-optimal performance from the initial stage. Specially, an expert can obtain the optimal scheduling policy by solving the formulated optimization problem with a few samples offline. For online learning, we train agent policies by following the expert’s demonstration with an acceptable performance gap in theory. Performance results show that our solution has a significant advantage with more than 50 percent improvement compared with the benchmark.

Age of Information (AoI) has become an important metric to evaluate the freshness of information, and studies of minimizing AoI in wireless networks have drawn extensive attention. In mobile edge networks, changes in critical levels for distinct information is important for users’ decision making, especially when merely partial observations are available. However, existing research has not yet addressed this issue, which is the subject of this paper. To address this issue, we first establish a system model, in which the information freshness is quantified by changes in its critical levels. We formulate Age-of-Critical-Information (AoCI) minimization as an optimization problem, with the purpose of minimizing the average relative AoCI of mobile clients to help them make timely decisions. Then, we propose an information-aware heuristic algorithm that can reach optimal performance with full obsevations in an offline manner. For online scheduling, an imitation learning-based scheduling approach is designed to choose update preferences for mobile clients under partial observations, where policies obtained by the above heuristic algorithm are utilized for expert policies. Finally, we demonstrate the superiority of our designed algorithm from both theoretical and experimental perspectives.

Tag identification is a critical operation in large-scale RFID applications. Typically, in the RFID-enabled warehouse, the reader needs to execute tag identifications to obtain the inventory information of numerous tagged items. The existing schemes usually divide the time frame into multiple slots and map each tag to one of them for replying its identity. This imposes serious tag collisions because two or more tags may be mapped to the same slot and their responses corrupt with each other. When tag collision happens, all the collided tags cannot be identified by the reader, which significantly increases the identification delay. To overcome the collision problem in the identification process, this paper proposes a Group Tag Identification (GTI) framework to identify grouped tags in both singleton and collision slots. The key novelty of GTI is in leveraging tag-correlation to identify grouped tags in the collision slots without any extra transmission overhead. The main challenge of this work is to overcome the communication and architectural limitations of RFID systems in the context of building ID and slot-correlation between tags. Extensive simulations show that GTI significantly reduces the identification delay by up to 40 percent when compared with the state-of-the-art dynamical frame slotted aloha schemes.

Recently, the boosting growth of computation-heavy applications raises great challenges for the Fifth Generation (5G) and future wireless networks. As responding, the hybrid edge and cloud computing (ECC) system has been expected as a promising solution to handle the increasing computational applications with low-latency and on-demand services of computation offloading, which requires new computing resource sharing and access control technology paradigms. This work establishes a software-defined networking (SDN) based architecture for edge/cloud computing services in 5G heterogeneous networks (HetNets), which can support efficient and on-demand computing resource management to optimize resource utilization and satisfy the time-varying computational tasks uploaded by user devices. In addition, resulting from the information incompleteness, we design an evolutionary game based service selection for users, which can model the replicator dynamics of service subscription. Based on this dynamic access model, a Stackelberg differential game based cloud computing resource sharing mechanism is proposed to facilitate the resource trading between the cloud computing service provider (CCP) and different edge computing service providers (ECPs). Then we derive the optimal pricing and allocation strategies of cloud computing resource based on the replicator dynamics of users’ service selection. These strategies can promise the maximum integral utilities to all computing service providers (CPs), meanwhile the user distribution can reach the evolutionary stable state at this Stackelberg equilibrium. Furthermore, simulation results validate the performance of the designed resource sharing mechanism, and reveal the convergence and equilibrium states of user selection, and computing resource pricing and allocation.

This article studies the PBFT-based sharded permissioned blockchain, which executes in either a local datacenter or a rented cloud platform. In such permissioned blockchain, the transaction (TX) assignment strategy could be malicious such that the network shards may possibly receive imbalanced transactions or even bursty-TX injection attacks. An imbalanced transaction assignment brings serious threats to the stability of the sharded blockchain. A stable sharded blockchain can ensure that each shard processes the arrived transactions timely. Since the system stability is closely related to the blockchain throughput, how to maintain a stable sharded blockchain becomes a challenge. To depict the transaction processing in each network shard, we adopt the Lyapunov Optimization framework. Exploitingdrift-plus-penalty(DPP) technique, we then propose an adaptive resource-allocation algorithm, which can yield the near-optimal solution for each network shard while the shard queues can also be stably maintained. We also rigorously analyze the theoretical boundaries of both the system objective and the queue length of shards. The numerical results show that the proposed algorithm can achieve a better balance between resource consumption and queue stability than other baselines. We particularly evaluate two representative cases of bursty-TX injection attacks, i.e., the continued attacks against all network shards and the drastic attacks against a single network shard. The evaluation results show that the DPP-based algorithm can well alleviate the imbalanced TX assignment, and simultaneously maintain high throughput while consuming fewer resources than other baselines.

Efficiently analyzing geo-distributed datasets is emerging as a major demand in a cloud-edge system. Since the datasets are often generated in closer proximity to end users, traditional works mainly focus on offloading proper tasks from those hotspot edges to the datacenter to decrease the overall completion time of submitted jobs in a one-shot manner. However, optimizing the completion time of current job alone is insufficient in a long-term scope since some datasets would be used multiple times. Instead, optimizing the data distribution is much more efficient and could directly benefit forthcoming jobs, although it may postpone the execution of current one. Unfortunately, due to the throwaway feature of data fetcher, existing data analytics systems fail to re-distribute corresponding data out of hotspot edges after the execution of data analytics. In order to minimize the overall completion time for a sequence of jobs as well as to guarantee the performance of current one, we propose to re-distribute the data along with task offloading, and formulate corresponding ε-bounded data-driven task scheduling problem over wide area network under the consideration of edge heterogeneity. We design an online schemarunData, which offloads proper tasks and related data via piggybacking to the datacenter based on delicately calculated probabilities. Through rigorous theoretical analysis,runData is proved concentrated on its optimum with high probability. We implementrunData based on Spark and HDFS. Both testbed results and trace-driven simulations show that run Data re-distributes proper data via piggybacking and achieves up to 37 percent reduction on average response time compared with state-of-the-art schemas.

In distributed training, workers collaboratively refine the global model parameters by pushing their updates to the Parameter Server and pulling fresher parameters for the next iteration. This introduces high communication costs for training at scale, and incurs unproductive waiting time for workers. To minimize the waiting time, existing approachesoverlap communication and computationfor deep neural networks. Yet, these techniques not only require the layer-by-layer model structures, but also need significant efforts in runtime profiling and hyperparameter tuning. To make the overlapping optimizationsimpleandgeneric, in this article, we propose a new Parameter Server framework. Our solutiondecouplesthe dependency between push and pull operations, and allows workers toeagerlypull the global parameters. This way, both push and pull operations can be easily overlapped with computations. Besides, the overlapping manner offers a different way to address the straggler problem, where the stale updates greatly retard the training process. In the new framework, with adequate information available to workers, they can explicitly modulate the learning rates for their updates. Thus, the global parameters can be less compromised by stale updates. We implement a prototype system in PyTorch and demonstrate its effectiveness on both CPU/GPU clusters. Experimental results show that our prototype saves up to 54% less time for each iteration and up to 37% fewer iterations for model convergence, achieving up to 2.86× speedup over widely-used synchronization schemes.

Edge intelligence has emerged as a prevalent enabling technology to support various intelligent applications. Along with the prosperity, it also raises great concern on the security and privacy since the edge servers are usually shared and untrusted. The security-sensitive code (i.e., the pre-trained model) and data may be easily stolen by malicious tenants, and even untrusted infrastructure providers. To this end, Software Guard Extensions (SGX) is proposed to provide an isolated Trust Execution Environment (TEE) for security and privacy guarantee. However, we find that running tasks in SGX suffer certain performance degradation due to the limited Enclave Page Cache (EPC) size. This further leads to frequent page swapping operations and the high enclave call overhead, which are also influenced by the task (i.e., DNN layer) dispatching and scheduling. To this end, in this paper, we designLasagna, as an SGX based secure DNN inference acceleration framework, which explores the layered-structure of DNN models to well balance the usage of the scarce EPC resources and the computation resources. Lasagna mainly consists of a global task balancer and a local task scheduler, responding for task dispatching across distributed edge servers and task scheduling in local server, respectively. We evaluate Lasagna over different well-known DNN models, and the results show that Lasagna effectively speeds up the inference performance by$1.11\times -1.51\times$.

Communication has been considered as a major bottleneck in large-scale decentralized training systems since participating nodes iteratively exchange large amounts of intermediate data with their neighbors. Although compression techniques like sparsification can significantly reduce the communication overhead in each iteration, errors caused by compression will be accumulated, resulting in a severely degraded convergence rate. Recently, the error compensation method for sparsification has been proposed in centralized training to tolerate the accumulated compression errors. However, the analog technique and the corresponding theory about its convergence in decentralized training are still unknown. To fill in the gap, we design a method named ECSD-SGD that significantly accelerates decentralized training via error-compensated sparsification. The novelty lies in that we identify the component of the exchanging information in each iteration (i.e., the sparsified model update) and make targeted error compensation over the component. Our thorough theoretical analysis shows that ECSD-SGD supports arbitrary sparsification ratio and achieves the same convergence rate as the non-sparsified decentralized training methods. We also conduct extensive experiments on multiple deep learning models to validate our theoretical findings. Results show that ECSD-SGD outperforms all the start-of-the-art sparsified methods in terms of both the convergence speed and the final generalization accuracy.

A novel paradigm named Wireless Powered Mobile Edge Computing (WP-MEC) emerges recently, which integrates Mobile Edge Computing (MEC) and Wireless Power Transfer (WPT) technologies. It enables mobile clients to both extend their computing capacities by task offloading, and charge from edge servers via energy transmission. Existing studies generally focus on the centralized design of task scheduling and energy charging in WP-MEC networks. To meet the decentralization requirement of the near-coming 6G network, we propose an online learning algorithm for computation offloading in WP-MEC networks with a distributed execution manner. Specifically, we first define the delay minimization problem by considering task deadline and energy constraints. Then, we transform it into a primal-dual optimization problem based on the Bellman equation. After that, we design a novel neural model that learns both offloading and time division decisions in each time slot to solve the formulated optimization problem. To train and execute the designed algorithm distributivity, we form multiple learning models decentralized on edge servers and they work coordinately to achieve parameter synchronization. At last, both theoretical and performance analyses show that the designed algorithm has significant advantages in comparison with other representative schemes.

Most of the existing FL systems focus on a data-parallel architecture where training data are partitioned by samples among several parties. In some real-life applications, however, partitioning by features is also of practical relevance and the number of features is usually unbalanced among parties. The corresponding learning framework is referred to as Vertical Federated Learning (VFL). Though some pioneering work focused on VFL, the convergence properties of VFL on unbalanced features, especially when parties conduct different numbers of local updates concerning heterogeneous computational capabilities are still unknown. In this article, we propose a new learning framework to improve the training efficiency of VFL on unbalanced features. Given the number of features and the computational capability owned by each party, our thorough theoretical analysis exhibits that the number of local updates conducted by each party has a great effect on the convergence rate and the computational complexity, both of which jointly determine the overall training efficiency in an interrelated and sophisticated way. Based on our theoretical findings, we formulate an optimization problem and derive the optimal solution by selecting an adaptive number of local training rounds for each party. Extensive experiments on various datasets and models demonstrate that our approach significantly improves the training efficiency of VFL.

It is always a challenging problem to deliver a huge volume of videos over the Internet. To meet the high bandwidth and stringent playback demand, one feasible solution is to cache video contents on edge servers based on predicted video popularity. Traditional caching algorithms (e.g., LRU, LFU) are too simple to capture the dynamics of video popularity, especially long-tailed videos. Recent learning-driven caching algorithms (e.g., DeepCache) show promising performance, however, such black-box approaches are lack of explainability and interpretability. Moreover, the parameter tuning requires a large number of historical records, which are difficult to obtain for videos with low popularity. In this paper, we optimize video caching at the edge using a white-box approach, which is highly efficient and also completely explainable. To accurately capture the evolution of video popularity, we develop a mathematical model calledHRSmodel, which is the combination of multiple point processes, including Hawkes’ self-exciting, reactive and self-correcting processes. The key advantage of the HRS model is its explainability, and much less number of model parameters. In addition, all its model parameters can be learned automatically through maximizing the Log-likelihood function constructed by past video request events. Next, we further design an online HRS-based video caching algorithm. To verify its effectiveness, we conduct a series of experiments using real video traces collected from Tencent Video, one of the largest online video providers in China. Experiment results demonstrate that our proposed algorithm outperforms the state-of-the-art algorithms, with 15.5% improvement on average in terms of cache hit rate under realistic settings.

Following the roadmap of carbon neutrality, wireless communication systems are upgrading to use green energy that comes from renewable sources, e.g., sun, tide, and wind. Due to the volatile arrival of green energy, the on-grid energy is used as a backup for a green coordinated multiple point system. In this work, a weighted sum-rate maximization problem in thegreencoordinated multiple point system is investigated by expecting non-positive consumption of the on-grid energy in the long term. Motivated by the capacity-achieving property and simple implementation, an online zero-forcing dirty paper precoder is proposed to update the precoding matrices by combining statistical learning with the Lyapunov learning technique. A tradeoff relation is theoretically established to show that the long-term weighted sum rate approaches the${\mathcal{ O}}(V)$-neighbor of optimal value while the long-term on-grid energy increases at a rate of${\mathcal{ O}}({\scriptstyle {}^{\scriptstyle \log ^{2}(V)}}\hspace {-0.224em}/\hspace {-0.112em}{\scriptstyle \sqrt {V}})$, where$V$is an introduced control parameter. Numerical results are used to verify the performance of the proposed online adaptive precoder.

Edge intelligence has already been widely regarded as a key enabling technology in a variety of domains. Along with the prosperity, increasing concern is raised on the security and privacy of intelligent applications. As these applications are usually deployed on shared and untrusted edge servers, malicious co-located attackers, or even untrustworthy infrastructure providers, may acquire highly security-sensitive data and code (i.e., the pre-trained model). Software Guard Extensions (SGX) provides an isolated Trust Execution Environment (TEE) for task security guarantee. However, we notice that DNN inference performance in SGX is severely affected by the limited enclave memory space due to the resultant frequent page swapping operations and the high enclave call overhead. To tackle this problem, we propose Lasagna, an SGX oriented DNN inference performance acceleration framework without compromising the task security. Lasagna consists of a local task scheduler and a global task balancer to optimize the system performance by exploring the layered-structure of DNN models. Our experiment results show that our layer-aware Lasagna effectively speeds up the well-known DNN inference in SGX by 1.31x-1.97x.

Modern online multiple object tracking (MOT) methods usually focus on two directions to improve tracking performance. One is to predict new positions in an incoming frame based on tracking information from previous frames, and the other is to enhance data association by generating more discriminative identity embeddings. Some works combined both directions within one framework but handled them as two individual tasks, thus gaining little mutual benefits. In this paper, we propose a novel unified model with synergy between position prediction and embedding association. The two tasks are linked by temporal-aware target attention and distractor attention, as well as identity-aware memory aggregation model. Specifically, the attention modules can make the prediction focus more on targets and less on distractors, therefore more reliable embeddings can be extracted accordingly for association. On the other hand, such reliable embeddings can boost identity-awareness through memory aggregation, hence strengthen attention modules and suppress drifts. In this way, the synergy between position prediction and embedding association is achieved, which leads to strong robustness to occlusions. Extensive experiments demonstrate the superiority of our proposed model over a wide range of existing methods on MOTChallenge benchmarks. Our code and models are publicly available at https://github.com/songguocode/TADAM.

In a large-scale sharded blockchain, transactions are processed by a number of parallel committees collaboratively. Thus, the blockchain throughput can be strongly boosted. A problem is that some groups of blockchain nodes consume large latency to form committees at the beginning of each epoch. Furthermore, the heterogeneous processing capabilities of different committees also result in unbalanced consensus latency. Such unbalanced two-phase latency brings a large cumulative age to the transactions waited in the final committee. Consequently, the blockchain throughput can be significantly degraded because of the large transaction's cumulative age. We believe that a good committee-scheduling strategy can reduce the cumulative age, and thus benefit the blockchain throughput. However, we have not yet found a committee-scheduling scheme that works for accelerating block formation in the context of blockchain sharding. To this end, this paper studies a fine-balanced tradeoff between the transaction's throughput and their cumulative age in a large-scale sharded blockchain. We formulate this tradeoff as a utility-maximization problem, which is proved NP-hard. To solve this problem, we propose an online distributed Stochastic-Exploration (SE) algorithm, which guarantees a near-optimal system utility. The theoretical convergence time of the proposed algorithm as well as the performance perturbation brought by the committee's failure are also analyzed rigorously. We then evaluate the proposed algorithm using the dataset of blockchain-sharding transactions. The simulation results demonstrate that the proposed SE algorithm shows an overwhelming better performance comparing with other baselines in terms of both system utility and the contributing degree while processing shard transactions.

Edge Learning is an emerging distributed machine learning in mobile edge network. Limited works have designed mechanisms to incentivize edge nodes to participate in edge learning. However, their mechanisms only consider myopia optimization on resource consumption, which results in the lack of learning algorithm performance guarantee and longterm sustainability. In this paper, we propose Chiron, an incentive-driven long-term mechanism for edge learning based on hierarchical deep reinforcement learning. First, our optimization goal combines learning-algorithms metric (i.e., model accuracy) with system metric (i.e., learning time, and resource consumption), which can improve edge learning quality under a fixed training budget. Second, we present a two-layer H-DRL design with exterior and inner agents to achieve both long-term and short-term optimization for edge learning, respectively. Finally, experiments on three different real-world datasets are conducted to demonstrate the superiority of our proposed approach. In particular, compared with the state-of-the-art methods under the same budget constraint, the final global model accuracy and time efficiency can be increased by 6.5 % and 39 %, respectively. Our implementation is available at https://github.com/Joey61Liuyi/Chiron.

New applications such as smart homes, autonomous vehicles are leading an increasing research topic of convolutional neural network (CNN) based inference on IoT edge devices. Unfortunately, this scenario meets a huge roadblock caused by the limited computing resources owned by these devices. One popular solution is to execute inference on an edge cluster with parallelization schemes instead of on a single device. However, the heterogeneous edge devices and varied neural layers bring challenges to this process. In this paper, we propose a pipelined cooperation scheme (PICO) to efficiently execute CNN inference for edge devices. Our goal is to maximize throughput by reducing redundant computing meanwhile to keep the inference latency under a certain value. PICO divides the neural layers and edge devices into several stages. The input data is fed into the first stage and the inference result is produced at the last stage. These stages compose an inference pipeline. The execution time of each stage is optimized to approach the maximum throughput as close as possible. We also implement an adaptive framework to choose the best inference scheme under different workloads. In our experiment with 8 RaspberryPi devices, the average inference latency can be reduced by$1.7\sim 6.5\times$under different workloads, and the throughput can be improved by$1.8\sim 6.2\times$under various network settings.

Container, as a light-weight virtualization technology with the advantages of continuous integration and easy deployment, has been widely adopted to support diverse microservices. At runtime, non-local container images need to be frequently pulled from remote registries to local servers, resulting in large pulling traffic and hence long startup time. A distinctive feature in container-based microservice, which has not been exploited, is that container images are in layered structure and some common base layers can be shared between co-located microservices. In this paper, we propose a layer sharing microservice deployment and image pulling strategy which explores the advantage of layer sharing to speedup microservice startup and lower image storage consumption. The problem is formulated into an Integer Linear Programming (ILP) form. An Accelerated Distributed Augmented Lagrangian (ADAL) based distributed algorithm executed cooperatively by registries and servers is proposed. Through extensive trace driven experiments, we validate the high efficiency of our ADAL based algorithm as it accelerates the microservice startup by 2.30 times in average and reduces the storage consumption by 55.33%.

Sharding can significantly improve the blockchain scalability, by dividing nodes into small groups called shards that can handle transactions in parallel. However, all existing sharding systems adopt complete sharding, i.e., shards are isolated. It raises additional overhead to guarantee the atomicity and consistency of cross-shard transactions and seriously degrades the sharding performance. In this paper, we present Pyramid, the first layered sharding blockchain system, in which some shards can store the full records of multiple shards thus the cross-shard transactions can be processed and validated in these shards internally. When committing cross-shard transactions, to achieve consistency among the related shards, a layered sharding consensus based on the collaboration among several shards is presented. Compared with complete sharding in which each cross-shard transaction is split into multiple sub-transactions and cost multiple consensus rounds to commit, the layered sharding consensus can commit cross-shard transactions in one round. Furthermore, the security, scalability, and performance of layered sharding with different sharding structures are theoretically analyzed. Finally, we implement a prototype for Pyramid and its evaluation results illustrate that compared with the state-of-the-art complete sharding systems, Pyramid can improve the transaction throughput by 2.95 times in a system with 17 shards and 3500 nodes.

Major manufacturers and retailers are increasingly using RFID systems in supply-chain scenarios, where theft of goods during transport typically causes significant economic losses for the consumer. Recent sample-based authentication methods attempt to use a small set of random sample tags to authenticate the integrity of the entire tag population, which significantly reduces the authentication time at the expense of slightly reduced reliability. The problem is that it still incurs extensive initialization overhead when writing the authentication information to all of the tags. This paper presents KTAuth, a lightweight integrity authentication approach to efficiently and reliably detect missing tags and counterfeit tags caused by stolen attacks. The competitive advantage of KTAuth is that it only requires writing the authentication information to a small set of deterministic key tags, offering a significant reduction in initialization costs. In addition, KTAuth strictly follows the C1G2 specifications and thus can be deployed on Commercial-Off-The-Shelf RFID systems. Furthermore, KTAuth proposes a novel authentication chain mechanism to verify the integrity of tags exclusively based on data stored on them. To evaluate the feasibility and deployability of KTAuth, we implemented a small-scale prototype system using mainstream RFID devices. Using the parameters achieved from the real experiments, we also conducted extensive simulations to evaluate the performance of KTAuth in large-scale RFID systems.

How to efficiently coordinate multiple readers to work together is critical for high throughput in RFID systems. Existing researchers focus on designing efficient reader scheduling strategies that arrange adjacent readers to work in different time to avoid signal collisions. However, the impact of unbalanced tag number of readers on tag read throughput is still very challenging. In RFID systems, the distribution of tags is usually variable and uneven, which makes the number of tags covered by each reader (i.e. the load of it) imbalanced. This imbalance leads to different execution time for readers: the heavy load readers take longer time to collect all tags, while the other readers that finish execution earlier have to wait for nothing. To avoid this useless waiting and improve the system throughput, this paper focuses on the load balancing problem of multiple readers. It is an NP-hard problem, for which we design heuristic algorithms that adjust readers’ interrogation regions according to designed strategies to efficiently balance their loads. The amazing advantage of our algorithm is that it can be adopted by almost all existing protocols in multi-reader systems, including the reader scheduling protocol, to improve system throughput. Extensive experiments demonstrate that our algorithm can significantly improve the throughput in various scenarios.

Bitcoin is currently the cryptocurrency with the largest market share. Many previous studies have explored the security of Bitcoin from the perspective of blockchain mining. Especially on the double-spending attacks (DSA), some state-of-the-art studies have proposed various analytical models, aiming to understand the insights behind the double-spending attacks. However, we believe that advanced versions of DSA can be developed to create new threats for the Bitcoin ecosystem. To this end, this paper mainly presents a new type of double-spending attack named Adaptive DSA in the context of the Bitcoin blockchain, and discloses the associated insights. In our analytical model, the double-spending attack is converted into a Markov Decision Process. We then exploit the Stochastic Dynamic Programming (SDP) approach to obtain the optimal attack strategies towards Adaptive DSA. Through the proposed analytical model and the disclosed insights behind Adaptive DSA, we aim to alert the Bitcoin ecosystem that the threat of double-spending attacks is still at a dangerous level.

In recent years, personalized federated learning (pFL) has attracted increasing attention for its potential in dealing with statistical heterogeneity among clients. However, the state-of-the-art pFL methods rely on model parameters aggregation at the server side, which require all models to have the same structure and size, and thus limits the application for more heterogeneous scenarios. To deal with such model constraints, we exploit the potentials of heterogeneous model settings and propose a novel training framework to employ personalized models for different clients. Specifically, we formulate the aggregation procedure in original pFL into a personalized group knowledge transfer training algorithm, namely, KT-pFL, which enables each client to maintain a personalized soft prediction at the server side to guide the others' local training. KT-pFL updates the personalized soft prediction of each client by a linear combination of all local soft predictions using a knowledge coefficient matrix, which can adaptively reinforce the collaboration among clients who own similar data distribution. Furthermore, to quantify the contributions of each client to others' personalized training, the knowledge coefficient matrix is parameterized so that it can be trained simultaneously with the models. The knowledge coefficient matrix and the model parameters are alternatively updated in each round following the gradient descent way. Extensive experiments on various datasets (EMNIST, Fashion_MNIST, CIFAR-10) are conducted under different settings (heterogeneous models and data distributions). It is demonstrated that the proposed framework is the first federated learning paradigm that realizes personalized model training via parameterized group knowledge transfer while achieving significant performance gain comparing with state-of-the-art algorithms.

Recently, the term “Internet of Things” (IoT) has elicited escalating attention. The flexibility, agility, and ubiquitous accessibility have encouraged the integration between machine learning (ML) with IoT. However, there are many challenges that present the key inhibitors in moving ML to the public solution, such as centralized training, poor training efficiency, and heavy computing capabilities requirements. Therefore, bringing learning intelligence to edge IoT nodes has been spotlighted for some researches. Meanwhile, how to govern the use of learning results efficiently, reliably, scalably, and safely is hampered by the heterogeneity and nonconfidence among IoT nodes. In this article, we propose a blockchain-based collective Q-learning (CQL) approach to address the above issues, where lightweight IoT nodes are used to train parts of learning layers, then employing blockchain to share learning results in a verifiable and permanent manner. We further improve the traditional Proof of Work (PoW). Instead of solving a meaningless puzzle, we regard the learning process in the IoT node as a piece of work. Accordingly, the winner is the IoT node with the minimum reduced percentage of the learning loss function, referred to as the Proof-of-Learning (PoL) consensus protocol. Specifically, in order to show how the CQL approach works, we use it to address a networking integrated cloud-edge-end resource allocation in IoT. The experimental results reveal the superior performance of the proposed scheme.

Radio-frequency identification (RFID) has been utilized in many applications, such as supply chain and stock management in supermarkets. RFID systems in such practical applications are inherently dynamic because tags may move in and out frequently. One important but challenging problem in such systems is how to estimate the number of dynamic tags fast and reliably. This article proposes effective solutions to this problem, which guarantee the accuracy of estimation and time efficiency. Especially, we want to simultaneously estimate the number of tags that moved out of the system (missing tags) and the number of tags that entered the system (unknown tags) in a specified time interval. We design a novel method called time slot reuse (TSR) that generates two logical frames corresponding to the two types of tags from only one physical frame. Based on TSR, we propose a protocol called SSR that can accurately estimate the number of dynamic tags by using the generated logic frames. However, the performance of SSR degrades significantly when the disparity between the number of the two types of tags is remarkable. We further propose an enhanced version of SSR (ESSR), which overcomes this drawback by partitioning the frame into ranges and mapping different types of tags into different ranges. Rigorous theoretical analysis is performed to tune parameters in SSR and ESSR to minimize the execution time. The simulation results demonstrate up to 80% improvement in time efficiency when compared with state-of-the-art solutions to the same problem.

Modern machine learning (ML) applications are often deployed in the cloud environment to exploit the computational power of clusters. However, this in-cloud computing scheme cannot satisfy the demands of emerging edge intelligence scenarios, including providing personalized models, protecting user privacy, adapting to real-time tasks, and saving resource cost. In order to conquer the limitations of conventional in-cloud computing, there comes the rise of on-device learning, which makes the end-to-end ML procedure totally on user devices, without unnecessary involvement of the cloud. In spite of the promising advantages of on-device learning, implementing a high-performance on-device learning system still faces with many severe challenges, such as insufficient user training data, backward propagation (BP) blocking, and limited peak processing speed. Observing the substantial improvement space in the implementation and acceleration of on-device learning systems, we intend to present a comprehensive analysis of the latest research progress and point out potential optimization directions from the system perspective. This survey presents a software and hardware synergy of on-device learning techniques, covering the scope of model-level neural network design, algorithm-level training optimization, and hardware-level instruction acceleration. We hope this survey could bring fruitful discussions and inspire the researchers to further promote the field of edge intelligence.

Unmanned aerial vehicle (UAV)-enabled mobile edge computing has been recognized as a promising technology to flexibly and efficiently handle computation-intensive and latency-sensitive tasks in the era of fifth generation (5G) and beyond. In this paper, we study the problem of Service Provisioning for UAV-enabled mobile edge computiNg (SPUN). Specifically, under task latency requirements and various resource constraints, we jointly optimize the service placement, UAV movement trajectory, task scheduling, and computation resource allocation, to minimize the overall energy consumption of all terrestrial user equipments (UEs). Due to the non-convexity of the SPUN problem as well as complex coupling among mixed integer variables, it is a non-convex mixed integer nonlinear programming (MINLP) problem. To solve this challenging problem, we propose two alternating optimization-based suboptimal solutions with different time complexities. In the first solution with relatively high complexity in the worst case, the joint service placement and task scheduling subproblem, and UAV trajectory subproblem are iteratively solved by the Branch and Bound (BnB) method and successive convex approximation (SCA), respectively, while the optimal solution to the computation resource allocation subproblem is efficiently obtained in the closed form. To avoid the high complexity caused by BnB, in the second solution, we propose a novel approximation algorithm based on relaxation and randomized rounding techniques for the joint service placement and task scheduling subproblem, while the other two subproblems are solved in the same way as that of the first solution. Extensive simulations demonstrate that the proposed solutions achieve significantly lower energy consumption of UEs compared to three benchmarks.

When running in Parameter Server (PS), the Distributed Stochastic Gradient Descent (D-SGD) incurs significant communication delays and huge communication overhead due to the model synchronization. Moreover, considering the heterogeneity of computational capability among workers, traditional synchronization modes incur under-utilization of computational resources because fast workers have to wait for slow ones finishing the computation. Although our previous work OSP can effectively solve these problems by overlapping the computation and communication procedures and allowing adaptive multiple local updates in distributed training, it causes the staleness problem brought by the overlap, yielding a performance degradation. In this paper, we propose a new method named LOSP by introducing local compensation to our previous synchronization mechanism, which mitigates adverse effects caused by the overlapping synchronization. We theoretically prove that LOSP (1) preserves the same convergence rate as the sequential SGD for non-convex problems, and (2) exhibits good scalability due to the linear speedup property with respect to both the number of workers and the average number of local updates. Evaluations show that LOSP significantly improves performance over the state-of-the-art ones in terms of both convergence accuracy and communication cost.

Vital signs (e.g., respiration rate or heartbeat rate) sensing is of great importance to implement pervasive in-home healthcare. Traditional vital signs monitoring approaches usually require users to wear some dedicated sensors. These approaches are intrusive and inconvenient to use, especially for elderly people. Some non-intrusive vital signs monitoring approaches based on wireless sensing have been proposed in recent years. However, these approaches require the target user to be in situ during the monitoring process, which greatly limits their utilization in practical scenarios where the target users usually move around. In this paper, we propose RF-RMM, an RFID-based approach to accurate and continuous respiration monitoring for mobile users. The major challenge in respiration monitoring for moving people is that the tiny body displacement caused by the user's respiration is overwhelmed by the user's entire body movement. To address this issue, we propose a novel approach that uses a pair of tags to eliminate the effect of the user's body movement. We fuse the data from the paired tags to cancel the effect of the user's entire body movement and retain only the displacement caused by the user's respiration. Another challenging issue in implementing RF-RMM is how to resolve the phase ambiguity problem when the target user moves around, which becomes more serious than in the static case. We propose a distance tracking algorithm to track the phase transition during the user's movement, according to which the phase ambiguity problem can be well handled. We implement RF-RMM on commercial RFID devices and conduct extensive real-world experiments to evaluate its performance. The results show that RF-RMM achieves accurate respiration rate monitoring with an average error of 0.54 BPM in estimating different users' respiration rate and an average relative error of less than 13% in estimating the user's individual breath length.

With the growing amount of data, the demand of big data storage significantly increases. Through the cloud center, data providers can conveniently share data stored in the center with others. However, one practically important problem in big data storage is privacy. During the sharing process, data is encrypted to be confidential and anonymous. Such operation can protect privacy from being leaked out. To satisfy the practical conditions, data tranmission with multi receivers is also considered. Furthermore, this paper proposes the notion of pre-authentication for the first time, i.e., only users with certain attributes that have already been authenticated can participate in the data transmission. The pre-authentication mechanism combines the advantages of proxy conditional re-encryption multi-sharing mechanism with the attribute-based authentication technique, thus achieving attributes authentication before re-encryption, and ensuring the security of the attributes and data. Finally this paper proves that the system can resist several attacks and the proposed pre-authentication mechanism could significantly enhance the system security level.

The parameter server architecture has shown promising performance advantages when handling deep learning (DL) applications. One crucial issue in this regard is the presence of stragglers, which significantly retards DL training progress. Previous solutions for solving stragglers may not fully exploit the computation resource of the cluster as evidenced by our experiments, especially in the heterogeneous environment. This motivates us to design a heterogeneity-aware parameter server paradigm that addresses stragglers and accelerates DL training from the perspective of computation parallelism. We introduce a novel methodology named straggler projection to give a comprehensive inspection of stragglers and reveal practical guidelines to solve this problem in two aspects: (1) controlling each worker's training speed via elastic training parallelism control and (2) transferring blocked tasks from stragglers to pioneers to fully utilize the computation resource. Following these guidelines, we propose the abstraction of parallelism as an infrastructure and design the Elastic-Parallelism Synchronous Parallel (EPSP) algorithm to handle distributed training and parameter synchronization, supporting both enforcedand slack-synchronization schemes. The whole idea has been implemented into a prototype called Falcon which effectively accelerates the DL training speed with the presence of stragglers. Evaluation under various benchmarks with baseline comparison demonstrates the superiority of our system. Specifically, Falcon reduces the training convergence time, by up to 61.83, 55.19, 38.92, and 23.68 percent shorter than FlexRR, Sync-opt, ConSGD, and DynSGD, respectively.

Network Function Virtualization (NFV) has attracted significant attention from both industry and academia as an important paradigm change in network service provisioning. Most existing studies on admissions of NFV-enabled requests focused on deploying dedicated Virtualized Network Function (VNF) instances to serve each individual request without exploring the sharing of VNF instances among multiple user requests. However, with ever-growing demands of user services, exclusive usages of VNF instances in most networks drastically degrade the network performance and largely under-utilize the VNF instance resources. In this paper, we jointly explore two different VNF instance scaling techniques, horizontal scaling and vertical scaling techniques, to improve the network throughput while minimizing the operational cost of the network, where horizontal scaling that migrates some existing VNF instances from their current locations to new locations to allow the VNF instances to be shared by multiple requests to reduce the resource consumption and operational cost of the network, and vertical scaling that instantiates new VNF instances to meet the demands of new request admissions if existing VNF instances sharing becomes more expensive or the end-to-end delay requirements of currently executing requests will be violated. We first propose a unified framework of maximizing the network throughput, by admitting as many as NFV-enabled requests while meeting the end-to-end delay requirements of the admitted requests. We then provide an Integer Linear Programming (ILP) solution for the problem when the problem size is small. Otherwise, we devise an efficient algorithm through a non-trivial reduction that reduces the problem to the minimum-weight feedback arc set problem and the generalized assignment problem (GAP). We finally conduct experiments to evaluate the performance of the proposed algorithm. Experimental results demonstrate that the proposed algorithm outperforms a baseline algorithm and achieves a performance on a par with its optimal ILP solution.

The potential for the wide-scale acceptance of the Industrial IoT is limited by a lack of automation, real-time monitoring, and connectedness. However, the future communication trend towards 5G is expected to bring greater benefits to IIoT infrastructures in terms of high-speed transmission and ultra-low latency. Furthermore, with emerging techniques such as millimeter-wave (mmWave), massive multiple-input multiple-output (MIMO), and machine-to-machine (M2M) communications, the coupling of IIoT and 5G will advance profoundly. Despite these advantages, 5G-envisioned IIoT ecosystems are expected to face other potential concerns such as trust, security, and privacy. Apart from this, the challenges related to data storage and processing and computational complexities will also draw significant attention. To address the above-mentioned challenges, it's important to analyze data in real-time. In this direction, transfer learning (TL) can be a revolutionary breakthrough. TL fosters greater explorations and experimentations, leading to innovations and greater productivity.

Zero-shot learning aims at recognizing unseen classes (no training example) with knowledge transferred from seen classes. This is typically achieved by exploiting a semantic feature space shared by both seen and unseen classes, i.e., attribute or word vector, as the bridge. One common practice in zero-shot learning is to train a projection between the visual and semantic feature spaces with labeled seen classes examples. When inferring, this learned projection is applied to unseen classes and recognizes the class labels by some metrics. However, the visual and semantic feature spaces are mutually independent and have quite different manifold structures. Under such a paradigm, most existing methods easily suffer from the domain shift problem and weaken the performance of zero-shot recognition. To address this issue, we propose a novel model called AMS-SFE. It considers the alignment of manifold structures by semantic feature expansion. Specifically, we build upon an autoencoder-based model to expand the semantic features from the visual inputs. Additionally, the expansion is jointly guided by an embedded manifold extracted from the visual feature space of the data. Our model is the first attempt to align both feature spaces by expanding semantic features and derives two benefits: first, we expand some auxiliary features that enhance the semantic feature space; second and more importantly, we implicitly align the manifold structures between the visual and semantic feature spaces; thus, the projection can be better trained and mitigate the domain shift problem. Extensive experiments show significant performance improvement, which verifies the effectiveness of our model.

The explosive growth of mobile devices promotes the prosperity of novel mobile applications, which can be realized by service offloading with the assistance of edge computing servers. However, due to limited computation and storage capabilities of a single server, long service latency hinders the continuous development of service offloading in mobile networks. By supporting multi-server cooperation, Pervasive Edge Computing (PEC) is promising to enable service migration in highly dynamic mobile networks. With the objective of maximizing the system utility, we formulate the optimization problem by jointly considering the constraints of server storage capability and service execution latency. To enable dynamic service placement, we first utilize Lyapunov optimization method to decompose the long-term optimization problem into a series of instant optimization problems. Then, a sample average approximation-based stochastic algorithm is proposed to approximate the future expected system utility. Afterwards, a distributed Markov approximation algorithm is utilized to determine the service placement configurations. Through theoretical analysis, the time complexity of our proposed algorithm is linear to the number of users, and the backlog queue of PEC servers is stable. Performance evaluations are conducted based on both synthetic and real trace-driven scenarios, with numerical results demonstrating the effectiveness of our proposed algorithm from various aspects.

Pervasive edge computing refers to one kind of edge computing that merely relies on edge devices with sensing, storage and communication abilities to realize peer-to-peer offloading without centralized management. Due to lack of unified coordination, users always pursue profits by maximizing their own utilities. However, on one hand, users may not make appropriate scheduling decisions based on their local observations. On the other hand, how to guarantee the fairness among different edge devices in the fully decentralized environment is rather challenging. To solve the above issues, we propose a decentrailized computation offloading algorithm with the purpose of minimizing average task completion time in the pervasive edge computing networks. We first derive a Nash equilibrium among devices by stochastic game theories based on the full observations of system states. After that, we design a traffic offloading algorithm based on partial observations by integrating general adversarial imitation learning. Multiple experts can provide demonstrations, so that devices can mimic the behaviors of corresponding experts by minimizing the gaps between the distributions of their observation-action pairs. At last, theoretical and performance results show that our solution has a significant advantage compared with other representative algorithms.

The parameter server (PS) paradigm has achieved great success in deploying large-scale distributed Deep Learning (DL) systems. However, these systems implicitly assume that the cluster is homogeneous and this assumption does not hold in many realworld cases. Although the previous efforts are paid to address heterogeneity, they mainly prioritize the contribution of fast workers and reduce the involvement of slow workers, resulting in the limitations of workload imbalance and computation inefficiency. We reveal that grouping workers into communities, an abstraction proposed by us, and handling parameter synchronization at the community level can conquer these limitations and accelerate the training convergence progress. The inspiration of community comes from our exploration of prior knowledge about the similarity between workers, which is often neglected by previous work. These observations motivate us to propose a new synchronization mechanism named Community-aware Synchronous Parallel (CASP), which uses the Asynchronous Advantage Actor-Critic (A3C)-based algorithm to intelligently determine community configuration and fully improve the synchronization performance. The whole idea has been implemented in a prototype system called Petrel that achieves a good balance between convergence efficiency and communication overhead. The evaluation under various benchmarks with multiple metrics and baseline comparison demonstrates the effectiveness of Petrel. Specifically, Petrel accelerates the training convergence speed by up to 1.87 x faster and reduces communication traffic by up to 26.85 percent, on average, over the non-community synchronization mechanisms.

The multi-interface networks are efficient infrastructures to deploy distributed Deep Learning (DL) tasks as the model gradients generated by each worker can be exchanged to others via different links in parallel. Although this decentralized parameter synchronization mechanism can reduce the time of gradient exchange, building a high-performance distributed DL architecture still requires the balance of communication efficiency and computational utilization, i.e., addressing the issues of traffic burst, data consistency, and programming convenience. To achieve this goal, we intend to asynchronously exchange gradient pieces without the central control in multi-interface networks. We propose the Piece-level Gradient Exchange and Multi-interface Collective Communication to handle parameter synchronization and traffic transmission, respectively. Specifically, we design the gradient sketch approach based on 8-bit uniform quantization to compress gradient tensors and introduce the colayerabstraction to better handle gradient partition, exchange and pipelining. Also, we provide general programming interfaces to capture the synchronization semantics and build the Gradient Exchange Index (GEI) data structures to make our approach online applicable. We implement our algorithms into a prototype system called Canary by using PyTorch-1.4.0. Experiments conducted in Alibaba Cloud demonstrate that Canary reduces 56.28 percent traffic on average and completes the training by up to 1.61x, 2.28x, and 2.84x faster than BML, Ako on PyTorch, and PS on TensorFlow, respectively.

Load balancing is a difficult online decision-making problem in the current network. Recently, with the development of the programmable data-plane, it is feasible to perform flexibly load balance directly inside the network. This in-network load balance scheme can quickly adapt to the volatility of network traffic. However, previous in-network solutions are largely relying on the manual process. Inspired by recent successes in applying machine learning in online control, automating the in-network load balance process is thus appealing. But as a distributed control system, it behooves us to ask the critical question: “Can the distributed switches learn globally optimal scheduling policy and still be deployed in a distributed fashion to allow rapid reaction in real-time?” To tackle this question, we adopt a centralized learning and distributed execution framework and propose a multi-agent actor-critic reinforcement learning algorithm in this paper. The centralized “critic” is reinforced with the global network state and joint actions of all agents to ease the training process whilst distributed switches can take actions relaying on their local observations. In addition, a baseline scheme is introduced to solve the credit assignment problem in the multi-agent system. The extensive simulations are conducted to evaluate our proposed algorithm in comparison to state-of-the-art schemes.

The accurate identification of congested links is crucial for network performance monitoring. Network boolean tomography uses end-to-end path measurements to identify congested links, and appears as a significant alternative when direct link monitoring is not available. However, most of existing tomographic methods assume no correlations between links, i.e., the congestion of one link is assumed to be independent from the congestion of any others, hindering their applications in practice because links could become correlated during a joint optimization procedure of many network operations like traffic routing and balancing. In this paper, we study practical network boolean tomography without such an assumption. We elaborate on the ill-posed nature of network boolean tomography to highlight the significance of integrating link correlations, and model the congested link identification from end-to-end congestion observations of paths as a problem of Maximum A-Posteriori (MAP) estimation. To avoid the explicit acquisition of any priori knowledge of link correlations, we then propose a learning-based algorithm with Long Short-term Memory (LSTM), a special recurrent neural network that is good at learning statistical dependencies of sequence elements from historical data. Numerical results over real network topologies validate our learning-based network boolean tomography.

Compute First Networking (CFN) recently is proposed as an in-network computing paradigm for well balancing between the networking and computation resource scheduling. Thanks to the proliferation of network functions virtualization, the virtualized network functions can coexist with the computing services on a shared platform like edge computing environment. Thus, one critical issue incurred by CFN is how to manage and schedule the resources among various services from different over-the-top service provider (OSP) with different resource requirements, i.e., network slicing. In this paper, we first formulate the network slicing problem as a Stackelberg game problem and prove that there exists a Nash equilibrium beneficial to both the Network Slice Broker (NSB) and OSP. Furthermore, we propose a cooperative game model on the networking and computation resource allocation within each slice and invent a Nash bargaining solution to resolve the intra-slice resource competition for slice performance promotion. Simulation results are provided to validate the effectiveness and high efficiency of the our proposed game based network slicing and resource scheduling algorithm.

Recently, deep learning (DL) becomes increasingly important since its transformative effect on a wide range of applications. During inference process, the DL model is deployed on the cloud to answer online queries. One crucial issue in the progress of DL inference is energy consumption, which significantly retards computation performance. Therefore, many previous investigations decrease the energy consumption via computation reuse technique based on similarity. However, if input data consists individually from mobile devices, applying these schemes will significantly decline computation performance. Because in disordered individual inputs, similarity for reuse is difficult to exploit directly. Results of initial experimental observations show that (1) individual input data also has high similarity for reuse, and (2) the total similarity during computation process has a relation with the characteristics of input data. This motivates us to design a reordering scheme to enhance similarity for computation reuse. Our main approaches are using statistical theory to predict the similarities among input data, and determining the execution sequence. Based on these approaches, we propose an effective input reordering scheme for computation reuse to save energy consumption. The evaluation under various benchmarks demonstrates that the reordering scheme significantly outperforms the previous schemes, for instance, the computation reuse is enhanced to $1.1 \times$ and the energy consumption is minimized to 40% according to the configuration of traditional computation reuse technique.

In isolated network domains, global trustworthiness (e.g., consistent network view) is critical to the multiple-domain business partners who aim to perform the trusted corporations depending on each isolated network view. However, to achieve such global trustworthiness across distributed network domains is a challenge. This is because when multiple-domain partners are required to exchange their local domain views with each other, it is difficult to ensure the data trustworthiness among them. In addition, the isolated domain view in each partner is prone to be destroyed by malicious falsification attacks. To this end, we propose a blockchain-based approach that can ensure the trustworthiness among multiple-party domains. In this paper, we mainly present the design and implementation of the proposed trustworthiness-protection system. A cloud-based prototype and a local testbed are developed based on Ethereum. Finally, experimental results demonstrate the effectiveness of the proposed prototype and testbed.

In this work, we present a multi-objective Virtual Data Center Embedding (VDCE) scheme for multi-domain cloud computing setups. The primary focus of the proposed scheme is on -Energy minimization, SLA assurance, and reduced energy Prices; and is named as ESP-driven VDCE. In the preliminary phase of this work, we formulate the proposed scheme as an optimization problem. However, due to the intractability of the formulated problem, its is remodelled and divided into three sub-problems (SP), i.e., data center identification, virtual machine mapping, and virtual link embedding. The output of one SP serves as an input to the next SP, such that the search space can be significantly narrowed. Finally, the proposed approach for VDCE is extensively validated against other algorithms. The obtained results indicate that the proposed ESP-driven VDCE approach achieves almost 7.6% more energy-aware embeddings with 11.5% higher SLA levels and approximately 23% lower energy expenses.

Payment channel networks (PCN) have demonstrated its significant advantages in improving the scalability of blockchain. However, the existing work of PCN leads to serious privacy leakage problem that intermediate nodes along a payment path can collude to obtain the payment amounts and payment receivers. To address this problem, we propose to move PCN-related modules into the Trusted Execution Environment (TEE) commonly available on modern CPUs, so that adversaries cannot access the critical payment information protected by TEE, even though they compromise the software (e.g., blockchain clients or operating system) outside of TEE. An additional challenge is that adversaries can still infer payment receivers by observing the pattern of message transmissions among nodes. To hide payment receivers, we further propose to send redundant transactions to pseudo receivers to confuse adversaries. A fast algorithm with provable approximation ratio has been proposed to maximize the level of privacy protection under the constraint of communication overhead. Both experiments on a small-scale testbed and large-scale simulations are conducted to evaluate our proposal. The results show that our proposed solution outperforms existing work significantly.

By connecting the maritime users to Internet, e.g., boats, ships, etc., it is possible to operate maritime sensing and informatics across seas and oceans. Such marine Internet of things (MIoT) is urging intelligent maritime applications, e.g., real-time vessel tracking, navigation safety, autonomous shipping, etc. Due to the bandwidth limitation of conventional marine channels, broadband communication is desired for these emerging applications. In this paper, we consider operating the TV white space (TVWS) spectrum in 700MHz to support the near-sea surface communication for MIoT terminals. To better utilize the TV channel capacity, we propose a proactive and efficient link adaptation (LA) scheme based on nonlinear autoregressive neural network (NARNN) time series prediction. Specifically, the historical signal samplings are used to predict the near-sea-surface channel link status for the next transmission slot, which is then used to select a proper modulation and coding scheme (MCS) for the next egress frame. We have conducted extensive simulations, and show that the average channel utility can achieve almost 85% of the optimal capacity. The proposed LA scheme can provide useful inspirations for applying data analytics to efficient and adaptive LA schemes for mobile Internet of things.

As to address the impact of heterogeneity in distributed Deep Learning (DL) systems, most previous approaches focus on prioritizing the contribution of fast workers and reducing the involvement of slow workers, incurring the limitations of workload imbalance and computation inefficiency. We reveal that grouping workers into communities, an abstraction proposed by us, and handling parameter synchronization in community level can conquer these limitations and accelerate the training convergence progress. The inspiration of community comes from our exploration of prior knowledge about the similarity between workers, which is often neglected by previous work. These observations motivate us to propose a new synchronization mechanism named Community-aware Synchronous Parallel (CSP), which uses the Asynchronous Advantage Actor-Critic (A3C), a Reinforcement Learning (RL) based algorithm, to intelligently determine community configuration and fully improve the synchronization performance. The whole idea has been implemented in a system called Petrel that achieves a good balance between convergence efficiency and communication overhead. The evaluation under different benchmarks demonstrates our approach can effectively accelerate the training convergence speed and reduce synchro-nization traffic.

To overcome the limitations on the scalability of current blockchain systems, sharding is widely considered as a promising solution that divides the network into multiple disjoint groups processing transactions in parallel to improve throughput while decreasing the overhead of communication, computation, and storage. However, most existing blockchain sharding systems adopt a static sharding policy that cannot efficiently deal with the dynamic environment in the blockchain system, i.e., joining and leaving of nodes, and malicious attack. This paper presents SkyChain, a novel dynamic sharding-based blockchain framework to achieve a good balance between performance and security without compromising scalability under the dynamic environment. We first propose an adaptive ledger protocol to guarantee that the ledgers can merge or split efficiently based on the dynamic sharding policy. Then, to optimize the sharding policy under dynamic environment with high dimensional system states, a deep reinforcement learning-based sharding approach has been proposed, the goals of which include: 1) building a framework to evaluate the blockchain sharding systems from the aspects of performance and security; 2) adjusting the re-sharding interval, shard number and block size to maintain a long-term balance of the system’s performance and security. Experimental results show that SkyChain can effectively improve the performance and security of the sharding system without compromising scalability under the dynamic environment in the blockchain system.

Federated learning is promising in enabling large-scale machine learning by massive mobile devices without exposing the raw data of users with strong privacy concerns. Existing work of federated learning struggles for accelerating the learning process, but ignores the energy efficiency that is critical for resource-constrained mobile devices. In this paper, we propose to improve the energy efficiency of federated learning by lowering CPU-cycle frequency of mobile devices who are faster in the training group. Since all the devices are synchronized by iterations, the federated learning speed is preserved as long as they complete the training before the slowest device in each iteration. Based on this idea, we formulate an optimization problem aiming to minimize the total system cost that is defined as a weighted sum of training time and energy consumption. Due to the hardness of nonlinear constraints and unawareness of network quality, we design an experience-driven algorithm based on the Deep Reinforcement Learning (DRL), which can converge to the near-optimal solution without knowledge of network quality. Experiments on a small-scale testbed and large-scale simulations are conducted to evaluate our proposed algorithm. The results show that it outperforms the start-of-the-art by 40% at most.

Satellite communication systems usually work in low signal-to-noise ratio (SNR) circumstances owning to the limited satellites' link budgets. Large doppler frequency offset in low-SNR satellite communication systems severely influences the performance of frequency synchronization, while the frequency offset correction still remains an open problem under low SNR condition especially for short burst transmission. To solve such a problem in satellite communications, we present a novel method named GP-MASO-MLE, which comprises a coarse correction based on the objective function with Gaussian Process (GP) search and a fine correction based on Maximum Likelihood Estimation (MLE) jointly with turbo iterations. Specifically, the proposed method is appropriate for non-data-aided frequency offset correction in satellite communication systems. Simulation results show that the proposed algorithm can approach to the bit error rate (BER) performance bound of ideal frequency offset correction within 0.1 dB, moreover, the proposed algorithm has lower computational complexity compared with traditional multi-step search algorithms.

Federated Learning (FL) takes advantage of distributed data to jointly train a global deep learning model on many clients, without revealing local data to the central server for privacy guarantee. However, due to the heterogeneity of the FL clients, some poor performance clients may become stragglers, impeding the global training process. It is desirable to offload these stragglers' tasks to some high performance servers, but this is at the risk of data privacy leakage. To mitigate such problem, we introduce the edge servers empowered by Trusted Execution Environment (TEE) to securely help the FL clients with poor performance. With the consideration of limited computation resource in TEE, we further investigate how to select the clients for help. Considering the time-varying processing capabilities on the FL clients, we propose an exploration-exploitation based client selection algorithm. Via evaluating our algorithm in a practical FL training task, the experiments show that the proposed algorithm indeed accelerate training process thanks to its efficient client selection.

One non-negligible flaw of the convolutional neural networks (CNNs) based single image super-resolution (SISR) models is that most of them are not able to restore high-resolution (HR) images containing sufficient high-frequency information. Worse still, as the depth of CNNs increases, the training easily suffers from the vanishing gradients. These problems hinder the effectiveness of CNNs in SISR. In this paper, we propose the Dual-view Attention Networks to alleviate these problems for SISR. Specifically, we propose the local aware (LA) and global aware (GA) attentions to deal with LR features in unequal manners, which can highlight the high-frequency components and discriminate each feature from LR images in the local and global views, respectively. Furthermore, the local attentive residual-dense (LARD) block that combines the LA attention with multiple residual and dense connections is proposed to fit a deeper yet easy to train architecture. The experimental results verified the effectiveness of our model compared with other state-of-the-art methods.

Tag searching is a fundamental problem for a variety of radio frequency identification (RFID) applications. Prior works focus on single group searching, which refers to determining which ones in a given set of tags exist in the system. In this paper, we propose PTS, a protocol that can perform fast Parallel Tag Searching for multiple users simultaneously. Different from prior works that have to execute k times to search k groups separately, PTS obtains searching results for all the k groups with only one-shot execution. PTS achieves high parallelism due to some novel designs. First, we develop a grouping filter that encodes the membership of tags in different groups, with which non-target tags can be efficiently filtered out for all the k groups simultaneously. Second, we design two new codes, grouping code and mapping code, with which the remaining tags can quickly verify and confirm which group they belong to. We theoretically analyze how to set optimal parameters for PTS to minimize the execution time and conduct extensive simulation experiments to evaluate its performance. Compared with the state-of-the-art solutions, PTS significantly improves time efficiency in multiple group searching scenarios (by a factor of up to 7.54X when k = 10) and achieves the same time efficiency in single group searching scenarios.

One of the enabling technologies of Edge Intelligence is the privacy preserving machine learning paradigm called Federated Learning (FL). However, communication inefficiency remains a key bottleneck in FL. To reduce node failures and device dropouts, the Hierarchical Federated Learning (HFL) framework has been proposed whereby cluster heads are designated to support the data owners through intermediate model aggregation. This decentralized learning approach reduces the reliance on a central controller, e.g., the model owner. However, the issues of resource allocation and incentive design are not well-studied in the HFL framework. In this paper, we consider a two-level resource allocation and incentive mechanism design problem. In the lower level, the cluster heads offer rewards in exchange of the data owners' participation, and the data owners are free to choose among any clusters to join. Specifically, we apply the evolutionary game theory to model the dynamics of the cluster selection process. In the upper level, given that each cluster head can choose to serve a model owner, the model owners have to compete for the services of the cluster head. As such, we propose a deep learning based auction mechanism to derive the valuation of each cluster head's services. The performance evaluation shows the uniqueness and stability of our proposed evolutionary game, as well as the revenue maximizing property of the deep learning based auction.

With the technical advancement of transportation electrification and Internet of vehicle, an increasing number of electric vehicles (EVs) and related infrastructures (e.g., service stations with both charging and communication services) are deployed in the intelligent highway systems. Not only can EVs enter the service station areas for charging, but they can also upload/download cached data at service stations to access multiple networking services. However, as EVs are operated individually with their unique travelling patterns, questions arise as how to incent EVs so that both energy and communication resources are optimally allocated. In this paper, we propose an online pricing mechanism of EV charging and data caching for service stations along the highway. First, we design an online reservation system at each EV to decide the best service station to park when the EV enters the highway. Furthermore, based on the variant power system status, an online pricing mechanism is devised to update the charging and caching price based on Q-learning, by which EVs can be motivated to arrive at the designated station for services. Finally, simulation results validate the effectiveness of the proposed scheme in improving the station's utility.

With the ubiquitous deployment of mobile sensors and smart devices, the scope of Internet of things (IoT) has extended to the space of mobile networks, where IoT terminals are moving around instead of being fixed in buildings, ground infrastructures, etc. In this paper, we consider such mobile Internet of things (MIoT), and propose an autonomous rate control (RC) scheme for the uplink transmission from MIoT terminals to access stations. A deep reinforcement learning (DRL) based approach is designed to capture the channel variations of the link and to improve the effectiveness of the rate selection for each egress frame. Extensive simulations are conducted for MIoT terminals including vehicles and UAVs and show significant throughput performance improvement comparing with traditional methods, as well as the robustness and scalability of the DRL-RC algorithm. The proposed DRL-RC can provide inspirations for efficient and scalable link adaptation schemes for MIoT terminals.

With significant and continuing advances in information and communication technologies, the Internet of Things (IoT) will play an increasingly important role in domains, such as healthcare, transportation, finance, and energy. In an IoT system, billions of devices (e.g., sensors, wearables, and smart appliances) are connected to the global network infrastructure, and one associated phenomenon is the generation of a large volume of data. Apart from data volume, the velocity, variety, and veracity of these data will pose a significant burden on conventional networking infrastructures. However, as sensor and fifth-generation (5G) cellular technologies advance, so will the pervasiveness of IoT deployment. Parallel to this trend, cloud computing has been integrated with IoT in order to address limitations in existing IoT networks (e.g., storage and computing resources), and examples include Google cloud dataflow and Amazon IoT. However, cloud-centric IoT solutions may not be suited for delay-sensitive and computationally intensive applications, for example, due to resource availability, end-to-end latency, bandwidth, etc. Increasingly, large-scale IoT deployments demand high connectivity, interoperability, and orchestration which are necessary for minimizing latency and maximizing throughput. This highlights the importance of a distributed computing platform that can support the interactions between IoT and cloud computing systems.

Nowadays, the widely used Internet-of-Things (IoT) mobile devices (MDs) generate huge volumes of data, which need analyzing and extracting accurate information in real time by compute-intensive deep learning (DL) inference tasks. Due to its multilayer structure, the deep neural network (DNN) is appropriate for the mobile-edge computing (MEC) environment, and the DL tasks can be offloaded to DNN partitions deployed in MEC servers (MECSs) for speed-up inference. In this article, we first assume the arrival process of DL tasks as Poisson distribution and develop a tandem queueing model to evaluate the end-to-end (E2E) inference delay of DL tasks in multiple DNN partitions. To minimize the E2E delay, we develop a joint optimization problem model of partition deployment and resource allocation in MECSs (JPDRA). Since the JPDRA is a mixed-integer nonlinear programming (MINLP) problem, we decompose the original problem into a computing resource allocation (CRA) problem with fixed partition deployment decision and a DNN partition deployment (DPD) problem that optimizes the optimal-delay function related to the CRA problem. Next, we design a CRA algorithm based on Markov approximation and a low-complexity DPD algorithm to obtain the near-optimal solution in the polynomial time. The simulation results demonstrate that the proposed algorithms are more efficient and can reduce the average E2E delay by 25.7% with better convergence performance.

With the widespread applications of the Internet of Things (IoT), e.g., smart city, business, healthcare, etc., the security of data and devices becomes a major concern. Although blockchain can effectively enhance the network security and achieve fault tolerance, the huge resource consumption and limited performance of data processing restrict its deployments in IoT scenarios. Observing the heterogeneity and resource constraints, we intend to make blockchain accommodate both wimpy and brawny IoT devices. In this article, we present Tornado, a high-performance blockchain system based on space-structured ledger and corresponding algorithms, to enable blockchain in IoT. Specifically, we first design a space-structured chain architecture with novel data structures for promoting the network scalability. To address the huge heterogeneity of IoT, a novel consensus mechanism named collaborative-proof of work is developed. Moreover, we propose the space-structured greedy heaviest-observed subtree (S2GHOST) protocol for improving the resource efficiency of IoT devices. Additionally, a dynamic weight assignment mechanism in S2GHOST contributes to reflect the trustworthiness of data and devices. Extensive experiments demonstrate that Tornado can achieve a maximum throughput of 3464.76 transactions per second. The optimizations of propagation latency and resource efficiency are 68.14% and 30.56%, respectively.

Drive-thru Internet has been considered as an effective Internet access method for Internet of Vehicles (IoV). Through the opportunistic vehicle-to-roadside WiFi connection, it can provide high throughput performance with low communication cost for IoV applications, such as intelligent transportation system, automotive infotainment, etc. However, its usability is highly affected by a fundamental issue called rate adaptation (RA), which is to adjust the modulation and coding rate to adapt to the dynamic wireless channel between the vehicle and the roadside access point (AP). Conventional WiFi RA schemes are designed for indoor or quasistatic scenarios and do not account for the channel variations in drive-thru Internet. In this article, we study the limitation of applying existing RA schemes in drive-thru Internet and propose a reinforcement learning (RL)-based RA scheme to capture the potential channel variation patterns and efficiently select the rate for every vehicle's egress frame. Simulation results demonstrate that the proposed RA scheme outperforms the existing schemes in network throughput and that the efficiency of the learning model can be generalized under various conditions. The proposed RA method can provide useful inspirations for designing robust and scalable link adaptation protocols in IoV.

Internet of Things (IoT) generates large amounts of data at the network edge. Machine learning models are often built on these data, to enable the detection, classification, and prediction of the future events. Due to network bandwidth, storage, and especially privacy concerns, it is often impossible to send all the IoT data to the data center for centralized model training. To address these issues, federated learning has been proposed to let nodes use the local data to train models, which are then aggregated to synthesize a global model. Most of the existing work has focused on designing learning algorithms with provable convergence time, but other issues, such as incentive mechanism, are unexplored. Although incentive mechanisms have been extensively studied in network and computation resource allocation, yet they cannot be applied to federated learning directly due to the unique challenges of information unsharing and difficulties of contribution evaluation. In this article, we study the incentive mechanism for federated learning to motivate edge nodes to contribute model training. Specifically, a deep reinforcement learning-based (DRL) incentive mechanism has been designed to determine the optimal pricing strategy for the parameter server and the optimal training strategies for edge nodes. Finally, numerical experiments have been implemented to evaluate the efficiency of the proposed DRL-based incentive mechanism.

Hosting virtualized network functions (VNF) has been regarded as an effective way to realize network function virtualization (NFV). Considering the cost diversity in cloud computing, from the perspective of service providers, it is significant to orchestrate the VNFs and schedule the traffic flows for network utility maximization (NUM) as it implies maximal revenue. However, traditional heuristic solutions based on optimization models usually follow some assumptions, limiting their applicability. Recent studies have shown that deep reinforcement learning (DRL) is a promising way to tackle such limitations. However, DRL agent training also suffers from slow convergence problem, especially with complex control problems. We notice that optimization models actually can be applied to accelerate the DRL training. Therefore, we are motivated to design a model-assisted DRL framework for VNF orchestration in this paper. Other than letting the agent blindly explore actions, the heuristic solutions are used to guide the training process. Based on such principle, the DRL framework is also redesigned accordingly. Experiment results validate the high efficiency of our model-assisted DRL framework as it not only converges 23× faster than traditional DRL algorithm, but also with higher performance at the same time.

Edge computing is a new paradigm to provide strong computing capability at the edge of pervasive radio access networks close to users. A critical research challenge of edge computing is to design an efficient offloading strategy to decide which tasks can be offloaded to edge servers with limited resources. Although many research efforts attempt to address this challenge, they need centralized control, which is not practical because users are rational individuals with interests to maximize their benefits. In this article, we study to design a decentralized algorithm for computation offloading, so that users can independently choose their offloading decisions. Game theory has been applied in the algorithm design. Different from existing work, we address the challenge that users may refuse to expose their information about network bandwidth and preference. Therefore, it requires that our solution should make the offloading decision without such knowledge. We formulate the problem as a partially observable Markov decision process (POMDP), which is solved by a policy gradient deep reinforcement learning (DRL) based approach. Extensive simulation results show that our proposal significantly outperforms existing solutions.

A modern Network Function Virtualization (NFV) service is usually expressed in a service chain that contains a list of ordered network functions, each can run in one or multiple virtual machines. Although lots of efforts have been devoted to service chain deployment, the researchers normally consider a simple model of network functions where different service chains have their own network functions no matter whether some of the network function appliances are interdependent. In this paper, we study the service chain deployment by exploiting two types of correlations between network functions: the Coordination Effect due to information exchanges among multiple VMs running the same network function, and the Traffic-Change Effect where the volume of outgoing traffic is not necessarily equal to the volume of its incoming traffic at each network function because of packet manipulations such as compression and encryption. These two effects have not been studied simultaneously in the context of service chaining. With theobjective to maximize the profit measured by the admitted traffic minus the implementation cost, we first formulate a joint service-function deployment and traffic scheduling (SUPER) problem that is proved to be NP-hard. We then devise an approximation algorithm based on the Markov approximation technique and analyze its theoretical bound on the convergence time. Simulation results show that the proposed algorithm outperforms two existing benchmark algorithms significantly.

MapReduce plays a critical role as a leading framework for big data analytics. In this paper, we consider a geo-distributed cloud architecture that provides MapReduce services based on the big data collected from end users all over the world. Existing work handles MapReduce jobs by a traditional computation-centric approach that all input data distributed in multiple clouds are aggregated to a virtual cluster that resides in a single cloud. Its poor efficiency and high cost for big data support motivate us to propose a novel data-centric architecture with three key techniques, namely, cross-cloud virtual cluster, data-centric job placement, and network coding based traffic routing. Our design leads to an optimization framework with the objective of minimizing both computation and transmission cost for running a set of MapReduce jobs in geo-distributed clouds. We further design a parallel algorithm by decomposing the original large-scale problem into several distributively solvable subproblems that are coordinated by a high-level master problem. Finally, we conduct real-world experiments and extensive simulations to show that our proposal significantly outperforms the existing works.

The papers in this special section examine artificial intelligence (AI)-driven developments in 5G mobile communications for industrial automation applications from a Big Data perspective. With the recent advances in information and communication technologies, industrial automation is expanding at a rapid pace. This transition is characterized by “Industry 4.0”, the fourth revolution in the field of manufacturing. Industry 4.0, also called as “Industrial Internet of Things (IIoT)” or “Smart Factories”, is a reflection of new industrial revolution that is not only interconnected, but also communicate, analyze, and use the information to create a more holistic and better connected ecosystem for the industries.

As the great prevalence of various Internet of Things (IoT) terminals, how to solve the problem of isolated information among different IoT platforms attracts attention from both academia and industry. It is necessary to establish a trusted access system to achieve secure authentication and collaborative sharing. Therefore, this article proposes a distributed and trusted authentication system based on blockchain and edge computing, aiming to improve authentication efficiency. This system consists of physical network layer, blockchain edge layer and blockchain network layer. Through the blockchain network, an optimized practical Byzantine fault tolerance consensus algorithm is designed to construct a consortium blockchain for storing authentication data and logs. It guarantees trusted authentication and achieves activity traceability of terminals. Furthermore, edge computing is applied in blockchain edge nodes, to provide name resolution and edge authentication service based on smart contracts. Meanwhile, an asymmetric cryptography is designed, to prevent connection between nodes and terminals from being attacked. And a caching strategy based on edge computing is proposed to improve hit ratio. Our proposed authentication mechanism is evaluated with respect to communication and computation costs. Simulation results show that the caching strategy outperforms existing edge computing strategies by 6%-12% in terms of average delay, and 8%-14% in hit ratio.

Satellite-based communication technology has gained much attention in the past few years, where satellites play mainly the supplementary roles as relay devices to terrestrial communication networks. Unlike previous work, we treat the low-earth-orbit (LEO) satellites as secure data storage mediums. We focus on data acquisition from a LEO satellite based data storage system (also referred to as the LEO based datacenters), which has been considered as a promising and secure paradigm on data storage. Under the LEO based datacenter architecture, one fundamental challenge is to deal with energy-efficient downloading from space to ground while maintaining the system stability. In this paper, we aim to maximize the amount of data admitted while minimizing the energy consumption, when downloading files from LEO based datacenters to meet user demands. To this end, we first formulate a novel optimization problem and develop an online scheduling framework. We then devise a novel coflow-like “Join the first K-shortest Queues (JKQ)” based job-dispatch strategy, which can significantly lower backlogs of queues residing in LEO satellites, thereby improving the system stability. We also analyze the optimality of the proposed approach and system stability. We finally evaluate the performance of the proposed algorithm through conducting emulator based simulations, based on real-world LEO constellation and user demand traces. The simulation results show that the proposed algorithm can dramatically lower the queue backlogs and achieve high energy efficiency.

Crowdsensing has been well recognized as a promising approach to enable large scale urban data collection. In a typical crowdsensing system, the task owner usually needs to provide incentives to the users (say participants) to encourage their participation. Among existing incentive mechanisms, posted pricing has been widely adopted because it is easy to implement while ensuring truthfulness and fairness. One critical challenge to the task owner is to set the right posted price to recruit a crowd with small total payment and reasonable sensing quality, i.e., posted pricing problem for robust crowdsensing. However, this fundamental problem remains largely open so far. In this paper, we model the robustness requirement over sensing data quality as chance constraints in an elegant manner, and study a series of chance constrained posted pricing problems in crowdsensing systems. Although some chance-constrained optimization techniques have been applied in the literature, they cannot provide any performance guarantees for their solutions. In this work, we propose a binary search based algorithm, and show that using this algorithm allows us to establish theoretical guarantees on its performance. Extensive numerical simulations demonstrate the effectiveness of our proposed algorithm.

Caching popular videos at mobile edge servers (MESs) has been confirmed as a promising method to improve mobile users (MUs) perceived quality of experience (QoE) and to alleviate the server load. However, with the multiple bitrate encoding techniques prevalently employed in modern streaming services, caching deployment is challenging for the following three facts: (1) cooperative caching should be explored for MUs located at overlapped coverage areas of MESs; (2) there exists tradeoff consideration for caching either high bitrate videos or high diversity videos; and (3) the relationship between MU perceived QoE and MU received bitrate, known as QoE function, varies in different services. Aiming to maximize the MU perceived QoE, we formulate the multiple bitrate video caching problem, and prove this problem is NP-hard for any given positive and strictly increasing QoE function. We then propose a polynomial complexity algorithm based on a general QoE function, which can achieve an approximate ratio arbitrarily close to 1/2. Specifically, for a linear QoE function, we explore useful property of optimal solutions, based on which more efficient algorithms are proposed. We demonstrate the effectiveness of our solutions via both theoretical analysis and extensive simulations.

Network Virtualization (NV) techniques allow multiple virtual network requests to beneficially share resources on the same substrate network, such as node computational resources and link bandwidth. As the most famous family member of NV techniques, virtual network embedding is capable of efficiently allocating the limited network resources to the users on the same substrate network. However, traditional heuristic virtual network embedding algorithms generally follow a static operating mechanism, which cannot adapt well to the dynamic network structures and environments, resulting in inferior nodes ranking and embedding strategies. Some reinforcement learning aided embedding algorithms have been conceived to dynamically update the decision-making strategies, while the node embedding of the same request is discretized and its continuity is ignored. To address this problem, a Continuous-Decision virtual network embedding scheme relying on Reinforcement Learning (CDRL) is proposed in our paper, which regards the node embedding of the same request as a time-series problem formulated by the classic seq2seq model. Moreover, two traditional heuristic embedding algorithms as well as the classic reinforcement learning aided embedding algorithm are used for benchmarking our prpposed CDRL algorithm. Finally, simulation results show that our proposed algorithm is superior to the other three algorithms in terms of long-term average revenue, revenue to cost and acceptance ratio.

Power-aware task scheduling on processors has been a research hotspot in computing systems. Given an application G containing a set N of tasks {n1,...,n|N|}, and a system containing a set U of processors {u1,..., u|U|}, the power-aware task scheduling generally refers to finding the appropriate processor and frequency for each task ni, so as to make sure that all the tasks can be finished efficiently and the overall energy consumption is guaranteed. In this article, we study the problem of minimizing the schedule length for energy consumption constrained parallel applications on heterogeneous computing systems, where the schedule length refers to the time interval between starting the first task and finishing the last task. For this problem, existing work adopts a policy that preassigns the minimum energy consumption for each unassigned task. Nevertheless, our analysis reveals that, such a pre-assignment policy could be unfair for the low priority tasks, and it may not achieve an optimistic schedule length. Thereby, we propose a new task scheduling algorithm that suggests a weight-based mechanism to preassign energy consumption for unassigned tasks, and we provide the rigorous proof to show its feasibility. Further, we show that this idea can be extended to solve reliability maximization problems with energy consumption constraint or with both deadline and energy consumption constraints, where the reliability refers to the probability of executing application G without failures, and the deadline constraint refers to the “allowable” maximum schedule length. We have conducted extensive experiments based on real parallel applications. The experimental results consistently demonstrate that our proposed algorithms can achieve favourable performance, compared to state-of-the-art algorithms.

Existing works adopt the Edge-Oriented Resource Allocation (EORA) scheme, in which edge nodes cache services and schedule user requests to distribute workloads over cloud and edge nodes, so as to achieve high-quality services and low latency. Unfortunately, EORA does not fully take into account the fact that service providers are sometimes independent from the edge operators with their own objectives. To deal with the conflict and cooperation between service providers and edge nodes, we devise a service-oriented resource allocation (SORA) scheme, where edge nodes and service providers adjust their resource allocations to provide requested services. We first prove that such resource allocation problem is NP-hard. We then propose a three-sided cyclic game (3CG) involving users, edge nodes, and service providers who make their individual decisions by choosing respectively high-quality services, high-value users, and cost-effective edge nodes for service deployment. Based on 3CG, we prove the existence and approximation ratio of pure-strategy Nash equilibriums (NEs). We also develop both centralized and distributed approximate algorithms for resource allocation. Finally, extensive experimental results validate the effectiveness and convergence of the proposed algorithms.

Recently, the emergence of blockchain has stirred great interests in the field of Internet of Things (IoT). However, numerous non-trivial problems in the current blockchain system prevent it from being used as a generic platform for large-scale services and applications in IoT. One notable drawback is the scalability problem. Lots of projects and researches have been done to solve this problem. Nevertheless, they do not consider different users' conditions, only using a single consensus protocol as the best fit one, as well as the IoT system is heavily constrained by computing and networking resources. In this article, we study a permissioned blockchain-based IoT architecture. In order to improve the scalability of the blockchain system and meet the needs of different users, we propose a service-oriented permissioned blockchain, where different consensus protocols are launched according to users' quality of service (QoS) requirements. Specially, we quantify a few popular consensus protocols. Additionally, we select block producers, which need a great number of computation resources, as well as dynamically allocate network bandwidth to the blockchain system. We formulate consensus protocols selection, block producers selection, and network bandwidth allocation as a joint optimization problem. We then use a dueling deep reinforcement learning approach to solve the problem. Simulation results demonstrate the effectiveness of our proposed scheme.

The tremendous increase of computation-heavy applications has posed great challenges in terms of enhanced service coverage and high-speed data processing in the Fifth Generation (5G) networks. As responding, the integrated fog and cloud computing (FCC) system has been expected as an efficient approach to support low-latency and on-demand computing services. This work considers the computing resource market in an FCC system operated by one cloud computing service provider (CCP) and multiple fog computing service providers (FCPs), in which the CCP shares its cloud computing resource among FCPs and itself to serve users with computational tasks. To facilitate the resource trading between the CCP and FCPs, a Stackelberg differential game based resource sharing mechanism is proposed. In this mechanism, performance discrepancy is introduced as a penalty factor to denote the mismatch between the resource supply and demand, which will encourage all computing providers (CPs) to make their trading decisions that can truthfully reflect their resource capacity and requirements. In addition, an evolutionary game based replicator dynamics is established to analyze the users' service selection among CPs. Based on the established hierarchical game framework, interactions between user selection and computing resource sharing are investigated. The performance of the designed resource sharing mechanism is validated in the simulations, which also reveal the convergence and equilibrium states of user selection, resource pricing and resource allocation.

Random access is a key technology in satellite communication, as a large number of machine- type communication (MTC) terminals accessing the satellite makes it difficult to guarantee the access quality. Irregular repetition slotted ALOHA (IRSA) is one random access protocol relying on transmitting irregular number of replicas in multiple time slots, achieving a peak throughput at 0.8 in practical implementations. However, the probability of sending a certain number of replicas stays the same when given degree distribution, without considering the effects of different loads, which means there are extra useless packets sent and brings power waste in IRSA. Therefore, enhanced irregular repetition slotted ALOHA (EIRSA) based on tracking degree distribution control (TDDC) algorithm is proposed in this paper with adaptive degree distribution adjustment scheme to reduce the number of replicas while maintaining the same access performance with adaptation. Simulation results show that proposed protocol can achieve higher performance at the same power level and it is adaptive to load change.

Parameter Server (PS) has been widely used to train a large amount of data on multiple machines in parallel. In parameter server, a critical problem is how to effectively schedule multiple training jobs to minimize the job completion time. Some existing work has proposed methods of setting the number of concurrent workers. However, they do not effectively consider the topology of GPU placement which affects the efficiency of communication. This paper proposes a novel resource-to-time model based on the number of workers and the topology of GPU placement. According to the model, we propose an algorithm called TOPO-PS particularly for topology problem in parameter servers. The algorithm achieves the placement strategy based on graph mapping algorithm. Evaluation under various algorithms evidences the superiority of our algorithm. TOPO-PS yields shorter job completion, by up to 53.48% of that of FIFO and 88.77% of OASIS.

The Vehicle-to-Vehicle (V2V) communication is critical in vehicular networks, besides the fact that it can enable the message sharing among neighboring vehicles, it can also help to relay neighbors' data content to remote destinations during the opportunistic V2V contacting period. The efficiency of V2V data pipe is affected drastically by the intermittent V2V connection, which requires a buffer queue to cache the data contents for random disruption and resuming of the transmission. In this paper, we propose a queueing theoretical framework to analysis the effectiveness of the opportunistic V2V communication, which captures the connection interruption due to the vehicle mobility and trades off between the V2V transmission effectiveness and average data service delay. An M/G/1/K queue model with service interruption is setup to analysis the relationship between how many data tasks can be fulfilled via V2V data pipe and the average time to accomplish the data service. We also consider two different resuming types, namely, resuming from beginning and resuming from break point. The proposed queueing framework is validated via simulation based on PTV VISSIM trace, which can provide guideline in design and development of future V2V communication protocols.

Enabling connected vehicles is becoming an essential demand for modern mobility world. To envision the Internet of Vehicles (IoVs) and better support the transmission of the explosive vehicular data, the Ultra High Frequency(UHF) spectrum resource in TV White Space (TVWS) band is re-utilized to provide cognitively wireless access for vehicle users. The TVWS access can provide high throughput and wide coverage due to its wide bandwidth and penetrability. However, there remains a critical issue for vehicles to dynamically change the link rate for the egress frames to adapt to the channel variance. In this paper, we investigate deep learning based rate adaptation (RA) scheme for the vehicle users accessing to the TVWS band. We utilize a series of the recent Signal-to-Noise (SNR) records, and modeled the RA as a Time Series Classification (TSC) problem, which is solved by Deep Learning (DL) models to classifies the collected SNR series to the optimal rate selection for the frame to be transmitted. We compared three different DL models and show that the performance of the DL based RAs outperform conventional RAs. Such results could provide insightful guidance for applying machine learning in the RA problem for wireless vehicular access.

In most recent years, zero-shot recognition (ZSR) has gained increasing attention in machine learning and image processing fields. It aims at recognizing unseen class instances with knowledge transferred from seen classes. This is typically achieved by exploiting a pre-defined semantic feature space (FS), i.e., semantic attributes or word vectors, as a bridge to transfer knowledge between seen and unseen classes. However, due to the absence of unseen classes during training, the conventional ZSR easily suffers from domain shift and hubness problems. In this paper, we propose a novel ZSR learning framework that can handle these two issues well by adaptively adjusting semantic FS. To the best of our knowledge, our work is the first to consider the adaptive adjustment of semantic FS in ZSR. Moreover, our solution can be formulated to a more efficient framework that significantly boosts the training. Extensive experiments show the remarkable performance improvement of our model compared with other existing methods.

Unsupervised learning is becoming more and more important recently. As one of its key components, the autoencoder (AE) aims to learn a latent feature representation of data which is more robust and discriminative. However, most AE based methods only focus on the reconstruction within the encoder-decoder phase, which ignores the inherent relation of data, i.e., statistical and geometrical dependence, and easily causes overfitting. In order to deal with this issue, we propose an Exclusivity Enhanced (EE) unsupervised feature learning approach to improve the conventional AE. To the best of our knowledge, our research is the first to utilize such exclusivity concept to cooperate with feature extraction within AE. Moreover, in this paper we also make some improvements to the stacked AE structure especially for the connection of different layers from decoders, this could be regarded as a weight initialization trial. The experimental results show that our proposed approach can achieve remarkable performance compared with other related methods.

The multibeam satellite system (MBSS) has great potential for mobile communications in 5G era due to its superiority in terms of extensive coverage, large capacity and real-time service. In order to integrate the resource allocation in multiple dimensions and maximize the system capacity of the MBSS, the scenario of a sparse network with dense users is selected to investigate the resource allocation method in time dimension, frequency dimension, space dimension and power dimension. We first propose a multilevel clustering algorithm and a cross-cluster grouping algorithm to realize the beam scheduling, by which the interference is reduced in time dimension and space dimension. Based on the beam scheduling scheme, we further explore the relationship between the system capacity and the resource allocation in frequency dimension and power dimension, where a joint power allocation and subchannel selection algorithm is proposed to optimize the spectral efficiency. Our simulation results show that the proposed multiple-dimension resource allocation method is superior to the existing methods in system capacity and convergence, which is not only applicable for the resource allocation in the MBSS but also provides an efficient approach to solve the coupling resource allocation problem.

Nowadays, more and more researchers are paying their attention to green routing. In this paper, we consider power consumption as a kind of QoS (quality of service) and apply a new learning-based approach for energy efficient transportation and QoS routing. Compared with traditional rule-based methods, the proposed method can learn additional information from the networks to improve routing performance, and have the flexibility to meet different QoS requirements. First, we propose a new identification of network nodes, namely node vectors, and a basic routing algorithm using node vectors is designed accordingly. Then, energy efficient transportation and QoS routing are proposed by adding QoS constraints into the routing decision. Link attributes such as power consumption, bandwidth and delay can be learned from these node vectors with neural networks. The learned link attributes together with the estimated distance can be used for routing decisions with QoS constraints. Simulation results show that the proposed method is reliable in routing tasks, and can achieve a remarkable performance when compared with the state-of-the-art work on the delay constrained least cost path (DCLC) problem.

Edge computing is an effective approach for resource provisioning at the network edge to host virtualized network functions (VNF). Considering the cost diversity in edge computing, from the perspective of service providers, it is significant to orchestrate the VNFs and schedule the traffic flows for network utility maximization (NUM) as it implies maximal revenue. However, traditional model-based optimization methods usually follow some assumptions and impose certain limitations. In this paper, inspired by the success of deep reinforcement learning in solving complicated control problems, we propose a deep deterministic policy gradients (DDPG) based algorithm. We first formulate the NUM problem with the consideration of end-to-end delays and various operation costs into a non-convex optimization problem and prove it to be NP-hard. We then redesign the exploration method and invent a dual replay buffer structure to customize the DDPG. Meanwhile, we also apply our formulation to guide our replay buffer update. Through extensive trace-driven experiments, we show the high efficiency of our customized DDPG based algorithm as it significantly outperforms both model-based methods and traditional non-customized DDPG based algorithm.

Predicting service demand of users based on the historical network traffic data in a large area has been of great importance towards better service management for geo-distributed systems, such as edge computing. Existing works relying on complete information of historical service demands cannot deal with real-world scenarios that a significant portion of historical data is not available because collecting such data comprehensively is either costly or impossible. Motivated by the challenge of data incompleteness for the demand prediction, in this paper, we propose a deep neural network model based on the encoder-decoder architecture consisting of an encoder, a converter, and a decoder. This model successfully extracts global features from incomplete historical data, predicts the future information in the feature space, and finally generates complete future demands of the whole area from future features. We devise an encoder with mixed 3D and 2D convolutional layers such that rich temporal and spatial features of service demands can still be captured even when historical data are partially missing. We embed the conditional generative adversarial networks (GANs) into our decoder that generates real demand distribution with magnified quality. Another major difference from the traditional encoder-decoder architecture is to establish a fine-grained association between the features of history and future by adding the converter in the middle of encoder and decoder. With the three components, our model can overcome the data incompleteness challenge and use the extracted global features to make an accurate prediction of service demands. We conduct experiments on a real-world dataset with demands of telecommunications services in an urban environment. The experimental results show that our model achieves much higher prediction accuracy than the state-of-the-art approaches when only incomplete historical data are available.

Managing edge resources is one of key issues in edge computing. Recent works of resource allocation in edge computing focus on service caching and request scheduling on edge nodes by distributing tasks over cloud and edge nodes, so as to achieve high-quality services and low latency. Unfortunately, most schemes do not pay enough attention on service providers which also have their own resources. To serve users, service providers cooperate with edge nodes and are sometimes independent of the edge operators with their own objectives. To deal with the cooperation and conflict among users, service providers and edge nodes, we propose a cyclic selection model to depict the resource allocation among users, edge nodes and service providers where they cooperate for completing user requests and compete for their own interest. Such resource allocation problem can be formulated as a non-linear integer programming which is very difficult to solve. Thus, we propose a three-sided cyclic game (3CG) involving users, edge nodes, and service providers who make their decisions: users select preferred services, service providers select cost-effective edge nodes, edge nodes select high-value users. 3CG is proved to have pure-strategy Nash equilibriums and an approximation ratio. We develop central and distributed approximate algorithms for resource allocation. The evaluation results of 3CG show the effectiveness and efficiency of the proposed algorithms.

Gradient descent algorithms are widely used in machine learning. In order to deal with huge volume of data, we consider the implementation of gradient descent algorithms in a distributed computing setting where multiple workers compute the gradient over some partial data and the master node aggregates their results to obtain the gradient over the whole data. However, its performance can be severely affected by straggler workers. Recently, some coding-based approaches are introduced to mitigate the straggler problem, but they are efficient only when the workers are homogeneous, i.e., having the same computation capabilities. In this paper, we consider that the workers are heterogenous which are common in modern distributed systems. We propose a novel heterogeneity-aware gradient coding scheme which can not only tolerate a predetermined number of stragglers but also fully utilize the computation capabilities of heterogenous workers. We show that this scheme is optimal when the computation capabilities of workers are estimated accurately. A variant of this scheme is further proposed to improve the performance when the estimations of the computation capabilities are not so accurate. We conduct our schemes for gradient descent based image classification on QingCloud clusters. Evaluation results show that our schemes can reduce the whole computation time by up to 3× compared with a state-of-the-art coding scheme.

Parameter server paradigm has shown great performance superiority for handling deep learning (DL) applications. One crucial issue in this regard is the presence of stragglers, which significantly retards DL training progress. Previous solutions for solving straggler may not fully exploit the computation capacity of a cluster as evidenced by our experiments. This motivates us to make an attempt at building a new parameter server architecture that mitigates and addresses stragglers in heterogeneous DL from the perspective of computation parallelism. We introduce a novel methodology named straggler projection to give a comprehensive inspection of stragglers and reveal practical guidelines for resolving this problem: (1) reducing straggler emergence frequency via elastic parallelism control and (2) transferring blocked tasks to pioneer workers for fully exploiting cluster computation capacity. Following the guidelines, we propose the abstraction of parallelism as an infrastructure and elaborate the Elastic-Parallelism Synchronous Parallel (EPSP) that supports both enforced-and slack-synchronization schemes. The whole idea has been implemented in a prototype called Falcon which efficiently accelerates the DL training progress with the presence of stragglers. Evaluation under various benchmarks with baseline comparison evidences the superiority of our system. Specifically, Falcon yields shorter convergence time, by up to 61.83%, 55.19%, 38.92% and 23.68% reduction over FlexRR, Sync-opt, ConSGD and DynSGD, respectively.

Zero-shot learning (ZSL) aims at recognizing unseen classes with knowledge transferred from seen classes. This is typically achieved by exploiting a semantic feature space (FS) shared by both seen and unseen classes, i.e., attributes or word vectors, as the bridge. However, due to the mutually disjoint of training (seen) and testing (unseen) data, existing ZSL methods easily and commonly suffer from the domain shift problem. To address this issue, we propose a novel model called AMS-SFE. It considers the Alignment of Manifold Structures by Semantic Feature Expansion. Specifically, we build up an autoencoder based model to expand the semantic features and joint with an alignment to an embedded manifold extracted from the visual FS of data. It is the first attempt to align these two FSs by way of expanding semantic features. Extensive experiments show the remarkable performance improvement of our model compared with other existing methods.

In Software Defined Networks (SDNs), one of the key tasks in control plane is the monitoring and measurement of the whole network. A typical SDN consists of a set of switches and a logically centralized controller responsible for network state monitoring and flow scheduling, to which low cost and efficient flow statistics collection plays an important role. However, existing flow statistics collection methods mainly focus on single-path routing and cannot accurately capture the flow statistics in the case of multi-path routing (MPR), which is widely used in modern networks. In this paper, we are motivated to propose a Multi-Path oriented Flow Statistics Collection (MFSC) strategy to minimize the total communication cost for flow statistics collection. The problem is first formulated into an integer linear programming (ILP) form. After analyzing the complexity of this problem, we present a relaxation-based algorithm with an approximation factor p, where p is the maximum number of switches passed by each flow. The experiment results demonstrate that our proposed algorithm can reduce the total communication cost by over 14% compared with the traditional solutions.

When running in Parameter Server (PS), the Distributed Stochastic Gradient Descent (SGD) incurs significant communication delays because after pushing their updates, computing nodes (workers) have to wait for the global model to be communicated back from the master in every iteration. In this paper, we devise a new synchronization parallel mechanism named overlap synchronization parallel (OSP), in which the waiting time is removed by conducting computation and communication in an overlapped manner. We theoretically prove that our mechanism could achieve the same convergence rate compared to the sequential SGD for non-convex problems. Evaluations show that our mechanism significantly improves performance over the state-of-the-art ones, e.g., by 4× for both AlexNet and ResNet18 in terms of convergence speed.

Contactless gesture recognition has emerged as a promising technique to enable diverse smart applications, e.g., novel human-machine interaction. Among others, gesture recognition based on radio frequency identification (RFID) is preferred due to its prevalent availability, low cost, and ease in deployment. However, current RFID-based gesture recognition approaches usually use profile template matching to distinguish different gestures, making them suffer from large recognition latency and fail to support real-time applications. In this paper, we propose a real-time and accurate contactless RFID-based gesture recognition approach called ReActor. ReActor uses machine learning rather than time-consuming profile template matching to distinguish different gestures, and thus achieves both very low recognition latency and high recognition accuracy. The major challenge of our approach is to determine a set of suitable attributes that can preserve the profile features of the signals related to different gestures. We combine two types of attributes in ReActor: the statistics of the signal profile that characterize coarse-grained features and the wavelet (transformation) coefficients of the signal profile that characterize fine-grained local features, both of which can be calculated fast. Experimental results demonstrate that ReActor can recognize a gesture with average latency less than 51ms, two orders of magnitude faster than state-of-the-art approaches based on profile template matching. Furthermore, ReActor also achieves higher recognition accuracy than previous works due to its optimized attribute set.

Localization plays a pivotal role in wireless sensor networks. Many range-based localization algorithms have been proposed for 2-D sensor networks or densely deployed 3-D sensor networks. However, range-based localization in sparse 3-D sensor networks is still a challenging problem, because the sparseness of the network makes it difficult to obtain a proper order of nodes to be sequentially localized. The patch-and-stitching localization strategy can conquer the sparseness problem in 2-D networks, but for 3-D networks it is still unknown how to uniquely merge two patches when there are not enough common nodes. In this paper, we solve this challenging problem by deriving the conditions under which two subnetworks can be uniquely merged. In the proposed approach, we treat the translation parameters as unknowns and form a set of equations with which the unknowns can be uniquely solved. The novelty of our algorithm also lies in that we exploit both common nodes and connecting edges among adjacent subnetworks to merge them, resulting in very high chances that two subnetworks can be merged. We conduct extensive simulation experiments to evaluate the performance of the proposed algorithm. The results show that the proposed algorithm could localize more than 90% of nodes in sparse 3-D networks with average node degree of 11 and anchor ratio of 5%, while the best existing solution can localize only 52% of nodes in the same situation.

Network congestion will result in significant performance degradation or even failures of many bandwidth-hungry Internet of Things (IoT) applications. Accurate and efficient congested link identification has become a foundational issue to IoT applications like self-driving cars, digital health, smart city, and so on. However, directly monitoring the massive number of interior links often introduces high operation cost or even is infeasible in practice, giving rise to indirect monitoring techniques like network Boolean tomography. Nevertheless, in many networks, the number of their interior links is larger than their end-to-end paths, making it very challenging for network Boolean tomography to find a determined solution. To resolve this issue, most of current methods try to utilize some prerequisites, such as the link congestion probabilities. While these probabilities might be hard or even unable to be obtained accurately in dynamical networks, limiting the practical deployment. In this paper, we are motivated to design a framework of congested link identification without any prerequisite or assumption. We first novelly model the congested link identification procedures as a Markov decision processes (MDPs), and then employ a reinforcement learning technology, i.e., Q-learning, to solve this MDP. The simulation results show that our proposed scheme can autonomously and efficiently explore the unknown network environment, and is able to achieve better adaptivity and correctness, without any prior knowledge comparing to existing methods.

The Internet of Things (IoT) is envisioned as a distributed network formed by many end devices, e.g., the motes of wireless sensor network (WSN). These important IoT end devices enable ubiquitous sensing of environments and provide reliable services for mission-critical applications. However, programs running on WSN devices are typically interrupt-driven and prone to interrupt-induced concurrency bugs, which are primarily caused by erroneous interleavings among interrupt procedure instances (IPIs) (namely, executions of interrupt processing logic). In this paper, we use a set of dynamic bug patterns to characterize the concurrency bugs due to buggy access-interleavings among IPIs to shared resources, including shared memory locations and shared communication channels. By matching the above bug patterns, a dynamic analysis approach called disclosing and locating concurrency bugs of interrupt-driven IoT programs based on dynamic bug patterns (Daemon) is proposed to automatically detect and locate concurrency bugs in WSN programs. A GUI tool of Daemon is developed. As the empirical studies exhibit, the tool can discover concurrency bugs effectively and locate the buggy source lines visually.

In the Internet of Things (IoT) community, wireless sensor network (WSN) is a key technique to enable ubiquitous sensing of environments and provide reliable services to applications. WSN programs, typically interrupt-driven, implement the functionalities via the collaboration of interrupt procedure instances (IPIs, namely executions of interrupt processing logic). However, due to the complicated concurrency model of WSN programs, the IPIs are interleaved intricately and the program behaviors are hard to predicate from the source codes. Thus, to improve the software quality of WSN programs, it is significant to disentangle the interleaved executions and develop various IPI-based program analysis techniques, including offline and online ones. As the common foundation of those techniques, a generic efficient and real-time algorithm to identify IPIs is urgently desired. However, the existing instance-identification approach cannot satisfy the desires. In this paper, we first formally define the concept of IPI. Next, we propose a generic IPI-identification algorithm, and prove its correctness, real-time, and efficiency. We also conduct comparison experiments to illustrate that our algorithm is more efficient than the existing one in terms of both time and space. As the theoretical analyses and empirical studies exhibit, our algorithm provides the groundwork for IPI-based analyses of WSN programs in IoT environment.

As an innovated and revolutionized technology, blockchain has been applied in many fields, such as cryptocurrency, food traceability, identity management, or even market prediction. To discover its great potential, both industry and academia have paid great attention to it and numerous researches have been conducted. Based on the literature and industry whitepapers, in this survey, we unroll and structure the blockchain related discoveries and scientific results in many aspects. Particularly, we classify blockchain technologies into four layers and carry out a comprehensive study on the consensus strategies, the network, and the applications of blockchain. Different blockchain applications are put into the corresponding categories based on the fields, especially in Internet of Things (IoT). When introducing each layer, we not only organize and summarize the related works, but also discuss the fundamental issues and future research directions. We hope this survey could shed some light on the research of blockchain and serve as a guide for further studies.

Radio frequency identification (RFID)-based localization and activity recognition have attracted much research attention recently. They rely on accurate measurements of signal features, e.g., phase and received signal strength (RSS), in line-of-sight (LOS) condition to estimate the location or activity status of the target objects. However, the LOS requirement might be frequently breached by obstacles between reader and tags in real deployed RFID systems. The resulting non-LOS (NLOS) signal will greatly reduce localization or activity recognition accuracy. How to filter out NLOS in the localization/activity recognition process is therefore practically important for guaranteeing accuracy. In this paper, we propose the first LOS/NLOS path recognition approach to differentiate the signals by LOS path from the ones by NLOS path. The proposed approach is both precise (with precision higher than 0.95) and real-time in nature (with recognition delay less than 400 ms) due to the following innovative designs. First, we design a new metric that can precisely distinguish LOS and NLOS paths by considering the joint variance of phase and RSS. Second, we propose an efficient method to mitigate the negative impacts of phase ambiguity on recognition precision. Third, we sample over a selected subset of channels and use only a handful of readings to perform LOS/NLOS path recognition, which greatly reduces the recognition delay without sacrificing precision. We conducted extensive experiments with commercial-off-the-shelf RFID devices. The results show that our approach achieves high precision and recall in all testing cases, with a precision of up to 0.969 and a recall of up to 0.991. Furthermore, our approach can also distinguish between different types of obstacles with an accuracy as high as 0.93.

Network function virtualization (NFV) decouples the traditional network functions from specific or proprietary hardware, such that virtualized network functions (VNFs) can run in software form. By exploring NFV, a consecutive set of VNFs can constitute a service function chain (SFC) to provide the network service. From the perspective of network service providers, how to maximize the network utility is always one of the major concerns. To this end, there are two main issues need to be considered at runtime: 1) how to handle the unpredictable network traffic burst? and 2) how to fairly allocate resources among various flows to satisfy different traffic demands? In this paper, we investigate a fairness-aware flow scheduling problem for network utility maximization, with joint consideration of resource allocation and rate control. Based on a discrete-time queuing model, we propose a low-complexity online-distributed algorithm using the Lyapunov optimization framework, which can achieve arbitrary optimal utility with different fairness levels by tuning the fairness bias parameter. We theoretically analyze the optimality of the algorithm and evaluate its efficiency by both simulation and testbed-based experiments.

Electricity price forecasting is a significant part of smart grid because it makes smart grid cost efficient. Nevertheless, existing methods for price forecasting may be difficult to handle with huge price data in the grid, since the redundancy from feature selection cannot be averted and an integrated infrastructure is also lacked for coordinating the procedures in electricity price forecasting. To solve such a problem, a novel electricity price forecasting model is developed. Specifically, three modules are integrated in the proposed model. First, by merging of Random Forest (RF) and Relief-F algorithm, we propose a hybrid feature selector based on Grey Correlation Analysis (GCA) to eliminate the feature redundancy. Second, an integration of Kernel function and Principle Component Analysis (KPCA) is used in feature extraction process to realize the dimensionality reduction. Finally, to forecast price classification, we put forward a differential evolution (DE) based Support Vector Machine (SVM) classifier. Our proposed electricity price forecasting model is realized via these three parts. Numerical results show that our proposal has superior performance than other methods.

Big data analytics in datacenters often involve scheduling of data-parallel jobs. Traditional scheduling techniques based on improving network resource utilization are subject to limited bandwidth in datacenter networks. To alleviate the shortage of bandwidth, some cluster frameworks employ techniques of traffic compression to reduce transmission consumption. However, they tackle scheduling in a coarse-grained manner at task level and do not perform well in terms of flow-level metrics due to high complexity. Fortunately, the abstraction of coflow pioneers a new perspective to facilitate scheduling efficiency. In this paper, we introduce a coflow compression mechanism to minimize the completion time in data-intensive applications. Due to the NP-hardness, we propose a heuristic algorithm called Fastest-Volume-Disposal-First (FVDF) to solve this problem. For online applicability, FVDF supports stage pipelining to accelerate scheduling and exploits recurrent neural networks (RNNs) to predict compression speed. Meanwhile, we build Swallow, an efficient scheduling system that implements our proposed algorithms. It minimizes coflow completion time (CCT) while guaranteeing resource conservation and starvation freedom. The results of both trace-driven simulations and real experiments show the superiority of our algorithm, over existing one. Specifically, Swallow speeds up CCT and job completion time (JCT) by up to 1.47χ and 1.66χ on average, respectively, over the SEBF in Varys, one of the most efficient coflow scheduling algorithms so far. Moreover, with coflow compression, Swallow reduces data traffic by up to 48.41 percent on average.

In the Service-Function-Chaining (SFC) enabled networks, various sophisticated policy-aware network functions, such as intrusion detection, access control and unified threat management, can be realized in either physical middleboxes or virtualized network function (VNF) appliances. In this paper, we study the service chaining towards the hybrid SFC clouds, where both physical appliances and VNF appliances provide services collaboratively. In such hybrid SFC networks, the challenge is how to efficiently steer the service chains for traffic demands while matching their individual policy chains concurrently such that a utility associated with the total admitted traffic rate and the induced overheads can be maximized. We find such problem has not been well solved so far. To this end, we devise a Markov Approximation (MA) based algorithm. The approximation property of the proposed algorithm is also proved. Extensive evaluation results show that the proposed MA algorithm can yield near-optimal solutions and outperform other benchmark algorithms significantly.

Multicasting is a fundamental functionality of many network applications, including online conferencing, event monitoring, video streaming, and so on. To ensure reliable, secure, and scalable multicasting, a service chain that consists of network functions (e.g., firewalls, intrusion detection systems, and transcoders) usually is associated with each multicast request. We refer to such a multicast request with service chain requirement as an network function virtualization (NFV)-enabled multicast request. In this paper, we study NFV-enabled multicasting in a software-defined network (SDN) with an aim to maximize network throughput while minimizing the implementation cost of admitted NFV-enabled multicast requests, subject to network resource capacity, where the implementation cost of a request consists of its computing resource consumption cost in servers and its network bandwidth consumption cost when routing and processing its data packets in the network. To this end, we first formulate two NFV-enabled multicasting problems with and without resource capacity constraints and one online NFV-enabled multicasting problem. We then devise two approximation algorithms with an approximation ratio of $2M$ for the NFV-enabled multicasting problems with and without resource capacity constraints, if the number of servers for implementing the service chain of each request is no greater than a constant $M$ (≥1). We also study dynamic admissions of NFV-enabled multicast requests without the knowledge of future request arrivals with the objective to maximize the network throughput, for which we propose an efficient heuristic, and for the special case of dynamic request admissions, we devise an online algorithm with a competitive ratio of $O(\log n)$ for it when $M=1$ , where $n$ is the number of nodes in the network. We finally evaluate the performance of the proposed algorithms through experimental simulations. Experimental results demonstrate that the proposed algorithms are promising and outperform existing heuristics.

In cyber-physical social systems (CPSS), a group of volunteers report data about the physical environment through their cyber devices and data aggregation is widely utilized. An important issue in data aggregation for CPSS is to protect users’ privacy. In this article, we use bitwise XOR and propose a bit-choosing algorithm to realize privacy-preserving min, k -th min, and percentile computation. By our algorithm, the aggregator can confirm whether a user’s data value is equal to certain value or within certain scale. Consequently, it is also possible to count the number of users satisfying given conditions. Our bit-choosing algorithm makes sure that the users send non-repetition replies to the aggregator to raise the aggregation accuracy. We analyze the communication cost and the achievable accuracy of our algorithm. Via performance comparison against existing protocols, the efficiency and accuracy of our algorithm are verified.

Emerging cyber technology permeates not only our working environment but also our daily life. We enjoy various cyber-enabled online services, e.g., using smartphones for navigation and friendships via social network services (SNS) and doing business through the cyberspace, where a large amount of human-centric data are produced in both the man–man and man–machine systems. This issue aims to enhance the next generation of human-centered social computing and provide relevant theoretical and algorithmic support for the application of social networks. Accepted articles included in this special issue cover topics on human-centric big data analysis, social influence analysis, data-driven interdisciplinary modeling and analysis, user behavior modeling and analysis, cyber-enabled and artificial intelligent (AI)-enhanced healthcare services, and so on. These contents may draft promising perspectives for future research and demonstrate a variety of issues and solutions related to social computing. With the AI-enhanced technical analysis, these human-centric data can be effectively used to better serve humanity.

In the past decade, network data communication has experienced a rapid growth, which has led to explosive congestion in heterogeneous networks. Moreover, the emerging industrial applications, such as automatic driving put forward higher requirements on both networks and devices. On the contrary, running computation-intensive industrial applications locally are constrained by the limited resources of devices. Correspondingly, fog computing has recently emerged to reduce the congestion of content-centric networks. It has proven to be a good way in industry and traffic for reducing network delay and processing time. In addition, device-to-device offloading is viewed as a promising paradigm to transmit network data in mobile environment, especially for autodriving vehicles. In this paper, jointly taking both the network traffic and computation workload of industrial traffic into consideration, we explore a fundamental tradeoff between energy consumption and service delay when provisioning mobile services in vehicular networks. In particular, when the available resource in mobile vehicles becomes a bottleneck, we propose a novel model to depict the users' willingness of contributing their resources to the public. We then formulate a cost minimization problem by exploiting the framework of Markov decision progress (MDP) and propose the dynamic reinforcement learning scheduling algorithm and the deep dynamic scheduling algorithm to solve the offloading decision problem. By adopting different mobile trajectory traces, we conduct extensive simulations to evaluate the performance of the proposed algorithms. The results show that our proposed algorithms outperform other benchmark schemes in the mobile edge networks.

Today's data center operators deploy network policies in both physical (e.g., middleboxes, switches) and virtualized (e.g., virtual machines on general purpose servers) network function boxes (NFBs), which reside in different points of the network, to exploit their efficiency and agility respectively. Nevertheless, such heterogeneity has resulted in a great number of independent network nodes that can dynamically generate and implement inconsistent and conflicting network policies, making correct policy implementation a difficult problem to solve. Since these nodes have varying capabilities, services running atop are also faced with profound performance unpredictability. In this paper, we propose a Heterogeneous netwOrk Policy Enforcement (HOPE) scheme to overcome these challenges. HOPE guarantees that network functions (NFs) that implement a policy chain are optimally placed onto heterogeneous NFBs such that the network cost of the policy is minimized. We first experimentally demonstrate that the processing capacity of NFBs is the dominant performance factor. This observation is then used to formulate the Heterogeneous Network Policy Placement problem, which is shown to be NP-Hard. To solve the problem efficiently, an online algorithm is proposed. Our experimental results demonstrate that HOPE achieves the same optimality as Branch-and-bound optimization but is 3 orders of magnitude more efficient.

The concept of the industrial Internet of things (IIoT) is being widely applied to service provisioning in many domains, including smart healthcare, intelligent transportation, autopilot, and the smart grid. However, because of the IIoT devices' limited onboard resources, supporting resource-intensive applications, such as 3D sensing, navigation, AI processing, and big-data analytics, remains a challenging task. In this paper, we study the multi-hop computation-offloading problem for the IIoT-edge-cloud computing model and adopt a game-theoretic approach to achieving Quality of service (QoS)-aware computation offloading in a distributed manner. First, we study the computation-offloading and communication-routing problems with the goal of minimizing each task's computation time and energy consumption, formulating the joint problem as a potential game in which the IIoT devices determine their computation-offloading strategies. Second, we apply a free-bound mechanism that can ensure a finite improvement path to a Nash equilibrium. Third, we propose a multi-hop cooperative-messaging mechanism and develop two QoS-aware distributed algorithms that can achieve the Nash equilibrium. Our simulation results show that our algorithms offer a stable performance gain for IIoT in various scenarios and scale well as the device size increases.

Traditional networks employ expensive dedicated hardware devices as middleboxes to implement Service Function Chains of user requests by steering data traffic along middleboxes in the service function chains before reaching their destinations. Network Function Virtualization (NFV) is a promising virtualization technique that implements network functions as pieces of software in servers or data centers. The integration of NFV and Software Defined Networking (SDN) further simplifies service function chain provisioning, making its implementation simpler and cheaper. In this paper, we consider dynamic admissions of delay-aware requests with service function chain requirements in a distributed cloud with the objective to maximize the profit collected by the service provider, assuming that the distributed cloud is an SDN that consists of data centers located at different geographical locations and electricity prices at different data centers are different. We first formulate this novel optimization problem as a dynamic profit maximization problem. We then show that the offline version of the problem is NP-hard and formulate an integer linear programming solution to it. We third propose an online heuristic for the problem. We also devise an online algorithm with a provable competitive ratio for a special case of the problem where the end-to-end delay requirement of each request is negligible. We finally evaluate the performance of the proposed algorithms through experimental simulations. The simulation results demonstrate that the proposed algorithms are promising.

Enterprise users at different geographic locations generate large-volume data that is stored at different geographic datacenters. These users may also perform big data analytics on the stored data to identify valuable information in order to make strategic decisions. However, it is well known that performing big data analytics on data in geographical-located datacenters usually is time-consuming and costly. In some delay-sensitive applications, the query result may become useless if answering a query takes too long time. Instead, sometimes users may only be interested in timely approximate rather than exact query results. When such approximate query evaluation is the case, applications must sacrifice timeliness to get more accurate evaluation results or tolerate evaluation result with a guaranteed error bound obtained from analyzing the samples of the data to meet their stringent timeline. In this paper, we study quality-of-service (QoS)-aware data replication and placement for approximate query evaluation of big data analytics in a distributed cloud, where the original (source) data of a query is distributed at different geo-distributed datacenters. We focus on the problems of placing data samples of the source data at some strategic datacenters to meet stringent query delay requirements of users, by exploring a non-trivial trade-off between the cost of query evaluation and the error bound of the evaluation result. We first propose an approximation algorithm with a provable approximation ratio for a single approximate query. We then develop an efficient heuristic algorithm for evaluating a set of approximate queries with the aim to minimize the evaluation cost while meeting the delay requirements of these queries. We finally demonstrate the effectiveness and efficiency of the proposed algorithms through both experimental simulations and implementations in a real test-bed, real datasets are employed. Experimental results show that the proposed algorithms are promising.

The emergence of edge computing has witnessed a fast-growing volume of data on edge devices belonging to different stakeholders which, however, cannot be shared among them due to the lack of the trust. By exploiting blockchain's non-repudiation and non-tampering properties that enable trust, we develop a blockchain-based big data sharing framework to support various applications across resource-limited edges. In particular, we devise a number of novel resource-efficient techniques for the framework: (1) the PoC (Proof-of-Collaboration) based consensus mechanism with low computation complexity which is especially beneficial to the edge devices with low computation capacity, (2) the blockchain transaction filtering and offloading scheme that can significantly reduce the storage overhead, and (3) new types of blockchain transaction (i.e., Express Transaction) and block (i.e., Hollow Block) to enhance the communication efficiency. Extensive experiments are conducted and the results demonstrate the superior performance of our proposal.

With the increasing computation and storage capabilities of mobile devices, the concept of fog computing was proposed to tackle the high communication delay inherent in cloud computing, and also improve the security to some extent. This paper concerns with the privacy issue inherent in the sustainable fog computing platform. However, there is no universal solution to the privacy problem in fog computing due to the device heterogeneity. In this paper, we proposed a differential privacy-based query model for sustainable fog computing supported data center. We designed a method that can quantify the quality of privacy preserving through rigorous mathematical proof. The proposed method uses the query model to capture the structure information of the sustainable fog computing supported data center, and the datasets for the query result are mapped to real vectors. Then, we implemented the differential privacy preserving by injecting Laplacian noise. The experiment results demonstrated that the proposed method can effectively resist various popular privacy attacks, and achieve relatively high data utility under the premise of better privacy preserving.

The video streaming technique and 4G LTE networks have been developing at a high speed, while the advancement of the battery technology is relatively slow. A lot of efforts have been made on understanding the power consumption in general 4G LTE networks, while little attention has been focused on reducing the power consumption of mobile devices for online video streaming in 4G LTE networks. In this paper, we attempt to develop efficient optimization techniques to address this problem. It is an important and also interesting problem, while there are many challenges needing to be alleviated (e.g., the uncertainty of users' behavior modes). To attack the challenges, we suggest a self-adaptive method that can allow us to adjust various parameters dynamically and efficiently, achieving a relatively small energy consumption. We give the rigorous theoretical analysis for the proposed method, and conduct the empirical study to validate its effectiveness.

202001 bcma

Network Function Virtualization (NFV) and Software-Defined Networking (SDN) have been envisioned an essential milestone in the evolution of communication networks. Their integration provides a more flexible and easier manageable software-based network environment that induces high expectations for reducing capital expenditures (CAPEX) and operational costs (OPEX) of network service providers. They also introduce technical challenges. One such challenge is to manage the placement of VNFs and to steer the data traffic of each NFV-enabled request through its specified network functions. In this paper, we opportunistically adopt the flexibility and cost-efficiency of VNF and SDN for NFV-enabled request admissions in SDNs and formulate profit maximization problems for static and dynamic NFV-enabled request admissions. We first provide an integer linear programming (ILP) solution to the problem in the static version if the problem size is small; otherwise, we devise a fast approximation algorithm with a provable approximation ratio for the static request admissions. We then propose an efficient online algorithm for dynamic request admissions, by leveraging VNF instance migrations and idle VNF releases back to the system. We finally evaluate the performance of the proposed algorithms through experimental simulations. Simulation results demonstrate that the proposed algorithms are very promising.

In the scenario of frequency coexistence between the GEO (geostationary) and NGEO (non-geostationary) satellite networks, the NGEO system should not incur harmful interference to the GEO system according to the policy of the Radio Regulations. Therefore, spectrum sensing as a promising solution is applied widely in this scenario. With the increasing number of NGEO satellites in the space, one NGEO system could be affected by other NGEO systems while sensing the signal from the GEO system. Given these preconditions, the cognitive radio (CR) scenario considered in this paper is that: the GEO system is regarded as the primary user, one NGEO system is regarded as the secondary user, while another NGEO system is regarded as the interfering user. Meanwhile, all the satellite systems are supposed to operate with more than one discrete transmit power levels which is practical and fits the concept of adaptive power control. In our context, we propose a spectrum strategy using hypothesis testing as well as maximum a posterior (MAP) to differentiate the GEO signal from the interfering NGEO and noise, and then identify the specific power level utilized by the GEO system. Moreover, we derive the closed-form expressions for threshold of verifying the status of the GEO, and for decision regions to determine its power level. Finally, extensive simulations are provided to verify the proposed studies.

Fuel cells are promising power sources for green data centers thanks to its high energy-efficiency, low greenhouse gas emissions and high reliability. However, fuel cells have a unique feature called limited load following, i.e., they are slow in adjusting power supply due to mechanical limitation of fuel delivery. When power demand of data centers suddenly grows, fuel cells would fail to provide sufficient power supply. On the other hand, fuel cells are slow to reduce its power supply when demand decreases, leading to energy waste. In this paper, we study to mitigate the impact of limited load following by associating a set of heterogeneous batteries with fuel cells. These batteries with different characteristics (e.g., capacity, charging and discharging rate) can power data centers when the energy supply of fuel cells is insufficient. They are charged by excessive power supply when demand decreases. Given future power demand, we formulate the energy management problem as a mixed-integer nonlinear programming. An online algorithm is designed to solve the problem without future knowledge. We conduct extensive simulations using real-world traces and results show that our proposed algorithm significantly outperforms existing solutions.