Image steganography has emerged as a promising channel for transmitting task assignments, authentication credentials, and scheduling metadata across distributed cloud nodes without raising suspicion. However, existing steganography methods primarily focus on robustness against attacks while overlooking the credibility of the extracted information. In distributed cloud environments, blindly accepting incorrect or manipulated data can propagate faults across nodes, trigger inconsistent system states, or even cause cascading service failures. To address this, we propose C2IS, a Credibility-Aware Coverless Image Steganography framework designed for distributed cloud services. C2IS enables cloud nodes not only to extract hidden information, but also to assess its credibility before execution or propagation. Specifically, we propose a polar harmonic transform-based feature extraction and selection strategy that extracts both coarse-grained and fine-grained features. The coarse-grained features, characterized by their high stability under various attacks, are utilized to carry secret information. Meanwhile, the fine-grained features quantify the degree of feature perturbation to assess information credibility. Furthermore, we develop a double median thresholding-based hash mapping algorithm, which binarizes inter-image feature differences using the median, thereby significantly enhancing the diversity of the generated hash sequences. Extensive experimental results show that our method achieves a complete hash sequence length of up to 13 bits, enabling higher hiding capacity. Meanwhile, under various attacks, it exhibits stronger robustness than state-of-the-art methods and can accurately determine whether the extracted information contains errors, providing a trustworthy means of covert communication for distributed cloud services.

Green Internet of Things (Green IoT) is a technique that intends to reduce energy consumption and carbon emissions of Internet of Things (IoT) devices by optimizing hardware design, communication protocols, and data processing. One of the most promising schemes to realize Green IoT is personalized Federated Learning (pFL). Unfortunately, existing pFL methods still need further improvement in achieving Green IoT from the following aspects. 1) Computational energy consumption: model training on IoT devices generates a substantial amount of computational energy consumption. 2) Model performance: the dynamic role differences in each layer of the trained deep neural network need to be considered. Jointly considering these aspects, we present a novel pFL framework named Energy-Efficient personalized Federated Learning (EE-pFL) for establishing Green IoT. Specifically, an IoT device serves as an edge server. Each IoT device produces a customized model through a model training phase and a model aggregation phase. In the model training phase, a threshold-based sparsification strategy is introduced to reduce the computational energy consumption of IoT devices by selectively executing parameter updates. In the model aggregation phase, layer aggregation and an Adaptive Weight Calculation (AWC) mechanism are proposed to capture dynamic role differences in different layers of a deep neural network. Experimental results demonstrate that EEpFL shows lower computational energy consumption and higher classification accuracy than advanced benchmarks.

The extensive local training in Hierarchical Federated Learning (HFL) imposes a substantial computational energy burden on end devices, a problem intensified by inherent system and data heterogeneity. While prior works attempt to mitigate this by using heterogeneous models or adjusting local training, they often suffer from critical drawbacks such as accuracy degradation from update biases and an inability to adapt to the dynamic nature of device resources. This paper introduces FedE2AD (Federated Energy-Efficient Adaptive Distillation), a novel cloud-network collaboration framework that leverages dynamic distillation to holistically optimize the energy-accuracy balance. At its core, FedE2AD implements this collaboration through a multi-level optimization approach. At the cloud layer, a Dynamic Model Allocation strategy intelligently assigns model architectures by assessing device status from static, dynamic, and data-centric perspectives. At the device layer, Variable Local Iterations enable real-time adaptation to fluctuating computational power. Crucially, to counteract model divergence, FedE2AD employs a Dual Knowledge Sharing mechanism at the edge layer, which uniquely combines direct aggregation of shared structures with data-free model distillation to ensure robust knowledge transfer. Experiments conducted on simulation platforms show that FedE2AD markedly outperforms existing methods. For instance, on the CIFAR-10 dataset under strong heterogeneity, it reduces single-round computation energy by 21.1% and increases final model accuracy by 1.42% compared to HDHRFL.

Large Language Model (LLM)-driven multi-agent task planning offers a promising approach for automating complex missions, particularly in critical domains like disaster response. Nevertheless, the efficacy of existing planners is often compromised in dynamic environments. This vulnerability primarily stems from their inability to enforce complex, non-linear task dependencies, coupled with a lack of mechanisms to recover from execution failures. To address these gaps in reliability and resilience, this paper introduces ReFlex-LLM, a novel cloud-edge framework for autonomic multi-UAV task planning. ReFlex-LLM leverages a Directed Acyclic Graph (DAG) to explicitly model and guarantee the logical correctness of task sequencing, thereby preventing foundational planning errors. In parallel, ReFlex-LLM incorporates a closed-loop feedback mechanism, allowing the central planner to receive execution status from edge UAVs and to adapt the mission plan in response to unforeseen contingencies. Extensive simulations on complex, dependency-heavy tasks demonstrate that ReFlex-LLM significantly outperforms state-of-the-art baselines, boosting the overall mission success rate by over 20%.

Extensive local training in Hierarchical Federated Learning (HFL) imposes high energy demands on resourceconstrained devices, a problem exacerbated by system heterogeneity which also causes performance variability. To address this, we propose HFHEMS, a novel method utilizing heterogeneous models and distillation to improve energy efficiency in HFL. HFHEMS employs dynamic model allocation to tailor computational loads to device capabilities and uses variable local iterations for real-time training adjustment. To preserve model accuracy, it integrates a dual knowledge sharing strategy with data-free distillation, enhancing knowledge transfer. Experiments confirm HFHEMS significantly reduces computation energy while maintaining robust accuracy, thus achieving a superior energy-accuracy trade-off.

Neural Architecture Search (NAS) can discover the optimal neural network architecture within a given SuperNet through automated search, which can improve model performance and reduce computational overhead on resource-constrained devices. Due to the vast SuperNet requiring substantial computational resources for training and evaluation, the search process is costly and difficult to apply directly on End Devices (EDs) with limited computational resources. Moreover, existing methods utilize zero-cost proxies to reduce computational costs in NAS, but overlook limited computational resources on EDs and waste a large amount of computational resources on the cloud server. Deploying NAS on the cloud server can effectively address this issue. The cloud server is used to search for the optimal Subnet, and EDs only need to train the Subnet based on local data. To this end, we propose a nonlinear aggregation-based Neural Architecture Search method based on Feature and Gradient zero-cost proxies (FG-NAS). Specifically, EDs upload local data characteristics to the cloud server, and then the cloud server uses FG-NAS to obtain an optimal SubNet model from the SuperNet based on the uploaded data characteristics. Finally, the cloud server sends the optimal SubNet to EDs, which can reduce the computational burden on EDs. Furthermore, FG-NAS evaluates the accuracy of neural architectures by considering both feature proxies during forward propagation and gradient proxies during backward propagation. Experiments on three datasets demonstrate that compared to current mainstream zero-cost proxy methods, FG-NAS can improve evaluation accuracy by an average of 1.04% and reduce single-network evaluation time by up to 2.45%.

Federated Learning (FL) can help multiple Internet of Things (IoT) devices to collaboratively train a machine learning model to provide intelligent services and applications (FL-AIoT). Due to IoT devices' limited storage capacity and energy, fresh data collected by devices often overwrites outdated data and establishes a heterogeneous data distribution. This causes the global model to forget outdated data's characteristics (i.e., catastrophic forgetting). Existing methods incorporate knowledge distillation into FL (i.e., federated distillation, FD) to extract and integrate characteristics from both fresh and outdated data, but they use fixed distillation temperatures for different devices, which overlooks that fixed distillation temperatures cannot match the heterogeneous data distribution on different devices and degrades global model accuracy. To this end, we propose a Federated Dynamic Decoupled Distillation method based on Logits distribution (Fed3DL). Specifically, Fed3DL utilizes decoupled distillation to mitigate catastrophic forgetting. To alleviate the impact of heterogeneous data distributions, Fed3DL novelly builds an adaptive temperature-aware mechanism to dynamically adjust the distillation temperature of each device based on the distribution of Logits. Additionally, Fed3DL introduces a regularization term into the local distillation loss to reduce inter-class characteristics disparity and improve model accuracy. Experiments on two datasets show that compared with the best of the 5 baselines, Fed3DL can improve the global model accuracy by an average of 3.40 %, reduce the forgetting rate by an average of 4.98 %, and achieve the lowest inter-class accuracy disparity.

In real-world Industrial Internet of Things (IIoT) scenarios, due to the limited storage capacity of IIoT devices, fresh data continuously received by diverse devices will overwrite the outdated data and change the local data distribution. However, state-of-the-art studies have demonstrated that federated learning tends to focus on training with fresh data, and the latest global model may forget the historical update directions (i.e., catastrophic forgetting). This issue can significantly degrade the global model accuracy. Existing methods primarily focus on integrating outdated data characteristics into fresh data but overlook the large parameter update gap between global and local models during global aggregation. This gap can cause the global model updates to deviate from the optimal direction. To this end, we propose a federated adaptive weighted aggregation method based on model consistency (FedAWAC). Specifically, FedAWAC measures the model consistency on devices and dynamically adjusts the aggregation weights of each local model, thereby guiding the global model toward optimal updates. Furthermore, FedAWAC integrates$\mathcal {M}$historical global models most correlated to the latest global model on the cloud server to overcome catastrophic forgetting. Experiments on four different datasets (nonidentically and independently distributed settings) indicate that compared to five baselines, FedAWAC can improve global model accuracy by an average of 1.86%, reduce the forgetting rate by an average of 3.91%, and save average memory usage by up to 2.57 GB.

Cloud-centric deployment of large models brings privacy concerns and communication burdens to IoT devices. Federated Learning (FL) can help numerous IoT devices collaboratively train a large model in a distributed manner to provide intelligent services and applications (AIoT). However, due to the limited storage capacity of IoT devices, fresh data collected by IoT devices often overwrites outdated data and establishes Non-IID data distributions. Moreover, state-of-the-art studies have indicated that FL tends to use fresh data for model training. This causes the global model to forget outdated data’s characteristics (i.e., catastrophic forgetting). Some methods utilize Knowledge Distillation (KD) to extract and integrate characteristics from both fresh data and outdated data. However, existing KD-based methods use fixed distillation temperatures for different IoT devices, which overlooks that fixed distillation temperatures cannot match the data distributions on different IoT devices and may degrade global model accuracy. To this end, we propose a Federated Dynamic Decoupled Distillation method based on Logits distribution (Fed3DL). Specifically, Fed3DL utilizes decoupled distillation to mitigate catastrophic forgetting. To dynamically adjust the distillation temperature of each IoT device, Fed3DL novelly builds an adaptive temperature-aware mechanism based on the Logits distribution. Furthermore, Fed3DL introduces a regularization term into the local distillation loss to improve global model accuracy. Experiments on four datasets show that Fed3DL can improve the global model accuracy by an average of 3.19%, reduce the forgetting rate by an average of 4.46%, and achieve the lowest inter-class accuracy disparity.

Previous state-of-the-art studies have demonstrated that adversaries can access sensitive user data by membership inference attacks (MIAs) in Federated Learning (FL). Introducing differential privacy (DP) into the FL framework is an effective way to enhance the privacy of FL. Nevertheless, in differentially private federated learning (DP-FL), local gradients become excessively sparse in certain training rounds. Especially when training with low privacy budgets, there is a risk of introducing excessive noise into clients’ gradients. This issue can lead to a significant degradation in the accuracy of the global model. Thus, how to balance the user's privacy and global model accuracy becomes a challenge in DP-FL. To this end, we propose an approach, known as differential privacy federated aggregation, based on significant gradient protection (DP-FedASGP). DP-FedASGP can mitigate excessive noises by protecting significant gradients and accelerate the convergence of the global model by calculating dynamic aggregation weights for gradients. Experimental results show that DP-FedASGP achieves comparable privacy protection effects to DP-FedAvg and cpSGD (communication-private SGD based on gradient quantization) but outperforms DP-FedSNLC (sparse noise based on clipping losses and privacy budget costs) and FedSMP (sparsified model perturbation). Furthermore, the average global test accuracy of DP-FedASGP across four datasets and three models is about$2.62$%,$4.71$%,$0.45$%, and$0.19$% higher than the above methods, respectively. These improvements indicate that DP-FedASGP is a promising approach for balancing the privacy and accuracy of DP-FL.

The Intelligent Autonomous Transport System in 6G (6G-IATS) refers to the coordination of 6G, Artificial Intelligence (AI), and intelligent transportation systems, which is expected to revolutionize future intelligent transportation systems. In 6G-IATS, large-scale traffic flow prediction, affiliated with time series prediction, holds significant value for transportation planning and urban management. As an emerging AI method, Large Language Models (LLMs) have emerged prominently in time series forecasting. Unfortunately, it is challenging to achieve accurate and efficient large-scale traffic flow prediction by LLMs in 6G-IATS, due to the two issues: a) these LLMs fail to capture the spatio-temporal correlations in a large-scale road network, leading to limited prediction accuracy, and b) they process a substantial amount of training data on the central server, which imposes low training efficiency. Jointly considering the two concerns, this paper proposes a novel LLM and edge computing-based architecture for large-scale traffic flow prediction in 6G-IATS, called Spatio-Temporal Generative Large Language Model on Edge (STGLLM-E). In this architecture, we first decompose the entire large-scale road network into several subgraphs. To capture the spatio-temporal correlations, an LLM-based method named Spatio-Temporal Generative Large Language Model (STGLLM) including Spatio-Temporal Module (STM) and Generative Large Language Model (GLLM) is proposed. Secondly, to improve the training efficiency of the STGLLM-E, an edge training strategy based on edge servers is devised. Experiments are conducted on two real-world traffic flow datasets. The experimental results illustrate that the STGLLM-E is superior to the baselines in the prediction accuracy and the efficiency of training.

Traffic speed prediction is significant for intelligent navigation and congestion alleviation. However, making accurate predictions is challenging due to three factors: 1) traffic diffusion, i.e., the spatial and temporal causality existing between the traffic conditions of multiple neighboring roads, 2) the poor interpretability of traffic data with complicated spatio-temporal correlations, and 3) the latent pattern of traffic speed fluctuations over time, such as morning and evening rush. Jointly considering these factors, in this paper, we present a novel architecture for traffic speed prediction, calledInterpretable Causal Spatio-Temporal Diffusion Network(ICST-DNET). Specifically, ICST-DNET consists of three parts, namely the Spatio-Temporal Causality Learning (STCL), Causal Graph Generation (CGG), and Speed Fluctuation Pattern Recognition (SFPR) modules. First, to model the traffic diffusion within road networks, an STCL module is proposed to capture both the temporal causality on each individual road and the spatial causality in each road pair. The CGG module is then developed based on STCL to enhance the interpretability of the traffic diffusion procedure from the temporal and spatial perspectives. Specifically, a time causality matrix is generated to explain the temporal causality between each road’s historical and future traffic conditions. For spatial causality, we utilize causal graphs to visualize the diffusion process in road pairs. Finally, to adapt to traffic speed fluctuations in different scenarios, we design a personalized SFPR module to select the historical timesteps with strong influences for learning the pattern of traffic speed fluctuations. Extensive experimental results on two real-world traffic datasets prove that ICST-DNET can outperform all existing baselines, as evidenced by the higher prediction accuracy, ability to explain causality, and adaptability to different scenarios.

In federated learning, multiple rounds of communication are involved between clients and the server to train a global model. The extensive model updates transmitted during the training lead to significant communication costs. Previous methods usually employ quantization or sparsification to compress model updates. However, the lossy compression leads to a decline in accuracy, it is challenging to strike a balance between communication efficiency and model accuracy. Meanwhile, due to the data heterogeneity, local updates among different clients are biased towards each other. Employing the same compression ratios for each local updates will further degrade the model accuracy. To achieve the trade-off between communication efficiency and model accuracy, we propose FedDAC, a Dual Adaptive Compression method in heterogeneous federated learning. In the local computation phase, the loss queue is adopted to detect the convergence trends within each client. FedDAC can then dynamically quantify model updates and allow for various compression ratios among heterogeneous clients. In the global aggregation phase, FedDAC can determine the fluctuations in training based on the similarity between clients and the server, thereby adjusting the sparsity ratio flexibly. To alleviate the reduction in model accuracy caused by lossy compression, we introduce residual updates in the local computation and global aggregation phases to maintain model accuracy. Experiment results show that compared with one-way compression methods NAGC and AdaQuantFL, FedDAC can maintain comparable accuracy while the accumulated communication volume is reduced by about 29.6 times, and 22.8 times, respectively. Moreover, the global model accuracy of FedDAC surpasses the two-way compression method T-FedAvg by about 2.4%, and the accumulated communication volume is about 2.5 times lower than T-FedAvg.

The rapid development of deep learning technologies and the widespread deployment of sensing devices have brought considerable attention to Internet of Things (IoT). The smart sensing application is one of the popular applications in IoT. Personalized Federated Learning (pFL) is a replacement to traditional Federated Learning (FL) to tackle the statistical heterogeneity of clients’ private datasets (e.g., non-Independent and Identically (non-IID) data). However, existing pFL methods encounter two challenges in smart sensing applications: a) the global model preference, causing poor global model performance for minority classes on sensing device data. b) the dynamic role differences in each hierarchy of the layer-stacked deep learning model that needs to be considered. Jointly considering these challenges, we present a novel edge-cloud enabled pFL framework named pFL-Sensing for smart sensing applications. Specifically, the sensing device serves as an edge server. Each edge server produces a customized model through two phases: a model training phase and a model aggregation phase. In the model training phase, we design a novel loss function to alleviate the issue of the global model preference. In the model aggregation phase, hierarchical aggregation and an Adaptive Weight Calculation (AWC) mechanism are proposed to capture the dynamic role differences of model hierarchies. We perform simulation experiments on real image and text classification tasks. Experimental results show that pFL-Sensing demonstrates higher classification accuracy than advanced pFL baselines.

Inspection plays a vital role in dam operations, facilitating safety assurance and environmental monitoring. As an emerging technique, multi-Unmanned Aerial Vehicle (UAV) coordination technique brings a promising prospect to dam inspection, in which the core of the technique is UAVs Coverage Path Panning (UAVs-CPP), i.e., designing flight routes for UAVs to ensure comprehensive coverage of all point of interests within a dam region. Unfortunately, the dam region inevitably encounters natural disasters such as, the thunderstorm, strong wind, and earthquake, which is called abnormal environment. The abnormal environment triggers two issues: 1) the natural disasters leads to communication obstacles between UAVs due to communication link interruption and signal loss, and 2) the coverage and energy consumption of UAVs cannot be neglected. The existing coverage path planning algorithms are insufficient in the abnormal environment. To tackle these issue, a novel Deep Reinforcement Learning (DRL)-based architecture is devised for UAVs-CPP in the dam inspection under the abnormal environment, called Trace Pheromone Multi-Agent Deep Deterministic Policy Gradient (TP-MADDPG). Specifically, we first integrate a trace pheromone mechanism and Multi-Agent Deep Deterministic Policy Gradient (MADDPG). The trace pheromone mechanism can simulate the natural pheromone, which promotes the inner indirect communication between UAVs. Furthermore, we employ a MADDPG to perform UAVs-CPP, in which a reward function is designed to conjointly optimize coverage and energy consumption. Finally, the experimental results demonstrate that TP-MADDPG shows significant UAVs-CPP performance contrasted with the advanced benchmarks.

In real Internet of Things (IoT) environments, IoT devices vary widely in data types and needs. IoT devices participating in Personalized Federated Learning (PFL) all have their own unique data characteristics, but there exist some similarities. Previous work enhances Personalized Local Models (PLMs)’ performance by clustering IoT devices’ PLM while ignoring dimensionality and communication volume, resulting in lower Global Model (GM) accuracy and PLMs’ performance. To this end, we propose a Personalized Federated Learning method based on Model Head Clustering (FedMHC). Specifically, FedMHC groups IoT devices with similar data characteristics and distributes different GMs to different device groups. FedMHC allows each IoT device to obtain a GM that best fits its local data characteristics and guides PLM training. FedMHC only clusters model head parameters on the server side. Thus, the edge server only needs to transmit the head parameters and a single shared feature extractor parameters during communication with IoT devices. The improvement can effectively address the issues of dimensionality and high communication volume. Experiments on CIFAR-100, Tiny-ImageNet, and AG News datasets demonstrate that FedMHC enhances the model accuracy by 1.79% and 5.9% in pathological heterogeneous scenarios, and by 1.43%, 0.92%, and 0.94% in practical heterogeneous scenarios, compared to the topperforming methods among 9 baselines.

Rule induction is an important approach for reasoning over Knowledge Graphs. Existing works mainly rely on searching for rule instances within the Knowledge Graph to induce rules. However, this approach may generate a vast search space, leading to inefficiency and difficulty in discovering rules that lack instance support. We propose a Logical Rule Induction based on Representation Learning (LRIRL) method, which can mine rules at the pattern level. By computing rule scores through vector representations of the rules, LRIRL can avoid the inefficiency caused by directly searching for rule instances in a vast search space. Furthermore, by incorporating the deductive nature of logical rules into the rule induction process, LRIRL can mine and evaluate rules even in the absence of rule instances. The generated rules can be used to perform more efficient and accurate reasoning tasks on the Knowledge Graph. Experimental results demonstrate that LRIRL outperforms baselines in both reasoning accuracy and rule mining efficiency on public datasets. Compared to the best baseline, LRIRL can achieve an average accuracy improvement of 3.43% in MRR, 1.40% in HITS@1, and 1.80% in HITS@10. Moreover, LRIRL can reduce rule mining time by an average of approximately 35% compared to the best baseline.

Knowledge Graph Reasoning based on Representation Learning maps the entities and relations into a vector space, assessing entity-pair similarities to deduce unknown facts. However, existing methods focus solely on the local importance of entities and ignore isolated entities, leading to the issue of missing feature information. To solve this, we propose Knowledge Representation Learning based on Global Information (KRLGI). KRLGI adopts an attention-based biased random walk algorithm to obtain global information and determine the importance of global entities. The global entity importance goes through a conversion into attention weights, and these weights are integrated with the local entity importance. Subsequently, the local and global entity importance can be used together to represent entity embeddings. Globally integrated representations can reveal richer semantics and enhance reasoning capabilities. Experimental results indicate that KRLGI outperforms other baselines in reasoning accuracy on four public datasets. Notably, on the FB15k-237 dataset, KRLGI shows significant improvements over the best baseline, with increases of 6.37% in MRR, 2.5% in HITS@1, and 5.5% in HITS@10.

The Internet of Things (IoT) has facilitated the generation of vast amounts of data, enabling advanced personalized healthcare services such as Human Activity Recognition (HAR) systems. Privacy concerns have driven the adoption of federated learning (FL) across multiple distributed healthcare devices. However, the lack of model adaptation for device-specific data, particularly in non-i.i.d. settings and the limited resource capabilities of these devices, presents an ongoing challenge for FL implementation on-devices. In this study, we introduce a new Personalized Orthogonal Low-Rank Adaptation (ProLoRA) method which provides efficient personalized HAR system. ProLoRA uses low-rank orthogonal transformations of the fully connected layers to mitigating interference with previously acquired personalized knowledge. Our method demonstrates superior personalized model performance and competitive global model accuracy while significantly reducing computational and memory overhead compared to existing state-of-the-art personalized FL techniques. Comprehensive empirical evaluations on HAR and PAMAP2 datasets validate the superior performance of ProLoRA in both accuracy and resource efficiency.

Abstract Scene text in natural images carries additional semantic information to aid in image classification. Existing methods lack full consideration of the deep understanding of the text and the visual text relationship, which results in the difficult to judge the semantic accuracy and the relevance of the visual text. This paper proposes image classification based on Cross modal Knowledge Learning of Scene Text (CKLST) method. CKLST consists of three stages: cross‐modal scene text recognition, text semantic enhancement, and visual‐text feature alignment. In the first stage, multi‐attention is used to extract features layer by layer, and a self‐mask‐based iterative correction strategy is utilized to improve the scene text recognition accuracy. In the second stage, knowledge features are extracted using external knowledge and are fused with text features to enhance text semantic information. In the third stage, CKLST realizes visual‐text feature alignment across attention mechanisms with a similarity matrix, thus the correlation between images and text can be captured to improve the accuracy of the image classification tasks. On Con‐Text dataset, Crowd Activity dataset, Drink Bottle dataset, and Synth Text dataset, CKLST can perform significantly better than other baselines on fine‐grained image classification, with improvements of 3.54%, 5.37%, 3.28%, and 2.81% over the best baseline in mAP, respectively.

Few-shot learning, which aims to identify new classes with few samples, is an increasingly popular and crucial research topic in the machine learning . Recently, the development of deep learning has deepened the network structure of a few-shot model, thereby obtaining deeper features from the samples. This trend led to an increasing number of few-shot learning models pursuing more complex structures and deeper features. However, discarding shallow features and blindly pursuing the depth of sample feature levels is not reasonable. The features at different levels of the sample have different information and characteristics. In this paper, we propose a few-shot image classification model based on deep and shallow feature fusion and a coarse-grained relationship score network (HFFCR). First, we utilize networks with different depth structures as feature extractors and then fuse the two kinds of sample features. The fused sample features collect sample information at different levels. Second, we condense the fused features into a coarse-grained prototype point. Prototype points can better represent the information in this class and improve classification efficiency. Finally, we construct a relationship score network, concatenating the prototype points and query samples into a feature map and sending it into the network to calculate the relationship score. The classification criteria for learnable relationship scores reflect the information difference between the two samples. Experiments on three datasets show that HFFCR has advanced performance.

The Internet of Medical Things (IoMT) stimulates the development of intelligent medical applications. As mental disorders become a global problem, emotion recognition has received widespread attention, as it can contribute to more comprehensive mental health monitoring and psychological assessment. Physiological signal-based emotion-aware monitoring is a particularly promising application due to its noninvasive and objective data collection. Recently, multimodal emotion recognition has been enhanced with wireless body area network (WBAN) access to IoMT, where wireless medical sensors are interconnected and abundant signals are acquired conveniently. However, how to synthesize these multisource physiological signals to facilitate emotion recognition is a challenging problem due to their heterogeneity and interference. To solve this problem, we propose a real-time differential multimodal transformer (Diff-MT), where the main components are the differential hyperinformation extraction (DHE) module, the multimodal global cross-attention encoder (MGCE), and the difference-augmented feature fusion (DFF). Ultimately, we endow the system with emotional awareness and distribute the state to IoMT devices. Extensive experiments demonstrate that the proposed Diff-MT exhibits superior performance compared to existing methods on the WESAD and DEAP datasets and is appropriate for IoMT-based healthcare.

Graph Convolutional Networks (GCNs) have been widely used in skeleton-based human action recognition and have achieved promising results. However, current GCN-based methods are limited by their inability to refine semantic-guided joint relations and perform adaptive multi-scale analysis. These limitations impair their performance, particularly for analogical actions involving the interaction of the same body parts (e.g., drinking water and eating) as well as deficient actions with limited spatial-temporal information (e.g., subtle action writing and transient action sneezing). To solve these problems, we propose Part-level Refined Spatial Graph Convolution (PR-SGC) and Scale-aware Temporal Graph Convolution (Sa-TGC) for optimal action representation. The PR-SGC divides the skeleton into body parts and embeds this high-level semantics to refine the physical adjacency matrix. The Sa-TGC leverages the dynamic scale-aware mechanism to extract context-dependent multi-scale features. On this basis, we develop a novel Scale-aware Graph Convolutional Network with Part-level Refinement (SaPR-GCN), which is on par with state-of-the-art benchmarks on NTU RGB+D 60, NTU RGB+D 120, and NW-UCLA datasets.

The goal of multiple-image hiding is to hide several secret images within another carrier image without significantly changing its appearance, and then perfectly reconstruct all of the secret images. The challenge is to ensure that the stego-image has great visual quality and can resist various steganalysis under the premise of hiding as much information as possible in one image. To address this issue, the majority of known image-hiding methods focus on hiding images using compression techniques. In this article, we present a novel multiple-image hiding method based on up-sampling and reversible color transformation. First, the interpolation algorithm up-samples the carrier image, so that the attribute of similar neighboring pixel values in the up-sampled image can significantly improve the effect of image hiding. The embedding procedure is then performed using the proposed Euclidean Distance (ED)-based block matching and reversible color transformation, which decreases the chance of local blurring in the stego-image. Experimental results show that the proposed method surpasses existing advanced methods by achieving an average of 33 dB and 28 dB of PSNR for the stego-image with a hiding capacity 2 BPP and 8 BPP, and obtaining 100% reconstructing accuracy for all secret images. It also has a high level of resistance to steganalysis and a strong robustness against various image-processing attacks.

As data sharing on the Internet of Medical Things (IoMT) become more complicated, the problems of divergent interests, unregulated policies, privacy and security, and the resource constraints of data owners have drawn the attention of researchers. To address the problems, this paper provides resource management in the IoMT using a proposed edge-empowered blockchain federated learning system. Also, an improved linear regressor model is proposed as the global learning model for the federated learning system. Gradient parameters are encrypted using Paillier encryption on the federated server side before they are shared by the federated clients. Blockchain is deployed to provide new security features for IoMT and edge computing. Moreover, all transactions of IoMT and edge devices are stored on the blockchain for secure cataloguing and auditing. Edge computing is employed to handle complex computing tasks on behalf of IoMT devices. Extensive simulations are conducted to validate the efficacy of the proposed system model. The results show that computing costs are minimized while still achieving the benefits of security and privacy in the proposed system. Furthermore, security analysis shows that the proposed system is protected from security attacks.

Graph Convolutional Networks (GCNs) have been widely used in skeleton-based human action recognition. In GCN-based methods, the spatio-temporal graph is fundamental for capturing motion patterns. However, existing approaches ignore the physical dependency and synchronized spatio-temporal correlations between joints, which limits the representation capability of GCNs. To solve these problems, we construct the directed diffusion graph for action modeling and introduce the activity partition strategy to optimize the weight sharing mechanism of graph convolution kernels. In addition, we present the spatio-temporal synchronization encoder to embed synchronized spatio-temporal semantics. Finally, we propose Directed Diffusion Graph Convolutional Network (DD-GCN) for action recognition, and the experiments on three public datasets: NTU-RGB+D, NTU-RGB+D 120, and NW-UCLA, demonstrate the state-of-the-art performance of our method.

Event causality identification is an important task in natural language processing. However, this task is highly challenging due to the high dependency of event context, text semantic ambiguity and insignificant causality features between text events. These issues lead to the low precision of causal relationship identification between events. We propose an Event Causality Identification Based on Feature Fusion (ECIFF) to improve the causality identification precision between events by integrating the context, semantics, and syntax of natural language. Firstly, we utilize BERT to capture the contextual features of events in natural language, enhancing the contextual embedding of events in different contexts. Secondly, based on an adversarial generative graph representation method, ECIFF learns a massive amount of causal relationships in the CauseNet, which can enhance the semantic representation of causes and effects of events. Next, we exploit the shortest dependency path to shorten the length of sentences and inductively learn all possible syntactic dependency relationships. Finally, the contextual, semantic and syntactic features are fused to synthetically determine the causal relationships among events. The experimental results indicate that our proposed approach significantly outperforms the state-of-the-art method LSIN: on the CTBank dataset, the precision, recall and F1-score of our approach are improved by 1.6%, 3.2% and 2.4%; on the ESL dataset, the precision, recall and F1-score of our approach are improved by 4.0%, 4.7% and 4.3%.

Causal relationships between events can reflect the historical evolution of events and provide an important reference for predicting future trends. Script prediction methods based on event graphs often struggle to adequately consider the complex interdependencies among events, leading to prediction biases. The Script Event Prediction Based on Causal Generalization Learning (SEPCG) method has been proposed to enhance the accuracy of script event prediction. SEPCG uses the graph attention network to learn the direct causal relationship similarity between known events and candidate events, the direct result event similarity between known events and candidate events, and utilizes double similarity generalization to learn the predicate type between known events and candidate events. SEPCG uses a Neural Tensor Network to learn parameter-level event embeddings and improve the model’s sensitivity to parameter-level changes. Finally, based on the generalized event embeddings, the BiLSTM network is used to simultaneously learn the forward contextual information from the known event to the candidate event direction, i.e., the cause to the result information, and the reverse contextual information from the candidate event to the known event direction, i.e., the result to the cause information. The BiLSTM is used to capture the temporal information of event chains at different levels. The effectiveness of the model is verified on the NYT dataset, with a 1.84% improvement in accuracy compared to the best baseline.

The great achievement of IoT and the wide use of edge devices have brought explosive growth in data. The quality and scale of data determine the performances of machine learning models. Federated learning has attracted widespread attention for its ability to use isolated data and protect data privacy. Models can represent excellent generalization capabilities through federated training. However, the large number of devices and complex models involved in federated training exacerbate the communication costs and degrade the performance of the global model. Although existing approaches can reduce communication costs, they ignore the degradation of global model accuracy in a heterogeneous environment. To alleviate the huge communication costs in federated learning, this paper focuses on reducing upstream and downstream communication frequency while ensuring global model accuracy. We propose a Two-way Delayed Updates method with Model Similarity in Communication-Efficient Federated Learning (FedTDMS). FedTDMS employs personalized local computation to improve global model accuracy on heterogeneous data. Combining 10-cal update relevance check and global model compensation, FedTDMS reduces the communication frequency in Federated Learning. We conduct experiments on the MNIST-FL and CFAR-10-FL datasets. Results show that FedTDMS can greatly optimize communication efficiency while maintaining good global model accuracy.

In federated learning with differential privacy, an obvious phenomenon of local gradient sparsification emerges in some training rounds. When training with low privacy budgets, there is a risk of excessive noise being introduced into the uploaded gradients, leading to a significant decrease in the accuracy of the global model. To tackle the trade-off between privacy protection and model accuracy with low privacy budgets, we propose a differential privacy federated aggregation method based on gradient sparsification (DP-FedAGS), which not only prevents excessive noise addition by protecting only significant gradients, but also accelerates global model convergence by dynamically calculating the weight of the gradient. Experimental results indicate that DP-FedAGS achieves comparable privacy protection to DP-FedAvg and cpSGD, while outperforming DPFedSNLC. Moreover, our approach respectively attains an approximate average test accuracy improvement of $2 .45 \%, 4 . 79$% and $0 . 29$% over the above three methods, rendering DP-FedAGS a promising approach for exploring a balance between privacy protection and model accuracy.

In the era of Information and Communication, big data is rapidly generated due to the increasing data-driven applications in Internet of Things (IoT). Effectively processing such data, e.g., knowledge learning, on resource-limited IoT becomes a challenge. In this paper, we introduce a digital twin-enabled IoT, in order to achieve hyper-connected experience, green communication, and sustainable computing. Although knowledge learning benefits from the proposed system, system latency and energy consumption are still our focus in the distributed learning architecture. To this end, we leverage Deep Reinforcement Learning (DRL) to present the deep deterministic policy gradient with double actors and double critics (D4PG) to manage the multi-dimensional resources, i.e., CPU cycles, DT models, and communication bandwidths, enhancing the exploration ability and improving the inaccurate value estimation of agents in continuous action spaces. Extensive experimental results prove that the proposed architecture can efficiently conduct knowledge learning, and our intelligent scheme can effectively improve the system efficiency.

Abstract State‐of‐the‐art object detection models rely on large‐scale datasets for training to achieve good precision. Without sufficient samples, the model can suffer from severe overfitting. Current explorations in few‐shot object detection are mainly divided into meta‐learning‐based methods and fine‐tuning‐based methods. However, existing models do not focus on how feature maps should be processed to present more accurate regions of interest (RoIs), leading to many non‐supporting RoIs. These non‐supporting RoIs can increase the burden of subsequent classification and even lead to misclassification. Additionally, catastrophic forgetting is inevitable in both few‐shot object detection models. Many models classify directly in low‐dimensional spaces due to insufficient resources, but this transformation of the data space can confuse some categories and lead to misclassification. To address these problems, the Feature Reconstruction Detector (FRDet) is proposed, a simple yet effective fine‐tune‐based approach for few‐shot object detection. FRDet includes a region proposal network (RPN) based on channel attention and space attention called Multi‐Attention RPN (MARPN) and a head based on feature reconstruction called Feature Reconstruction Head (FRHead). MARPN utilizes channel attention to suppress non‐supporting classes and spatial attention to enhance support classes based on Attention RPN, resulting in fewer but more accurate RoIs. Meanwhile, FRHead utilizes support features to reconstruct query RoI features through a closed‐form solution, allowing for a comprehensive and fine‐grained comparison. The model was validated on the PASCAL VOC, MS COCO, FSOD, and CUB200 datasets and achieved better results.

Abstract Dense video captioning aims to locate multiple events in an untrimmed video and generate captions for each event. Previous methods experienced difficulties in establishing the multimodal feature relationship between frames and captions, resulting in low accuracy of the generated captions. To address this problem, a novel Dense Video Captioning Model Based on Local Attention (DVCL) is proposed. DVCL employs a 2D temporal differential CNN to extract video features, followed by feature encoding using a deformable transformer that establishes the global feature dependence of the input sequence. Then DIoU and TIoU are incorporated into the event proposal match algorithm and evaluation algorithm during training, to yield more accurate event proposals and hence increase the quality of the captions. Furthermore, an LSTM based on local attention is designed to generate captions, enabling each word in the captions to correspond to the relevant frame. Extensive experimental results demonstrate the effectiveness of DVCL. On the ActivityNet Captions dataset, DVCL performs significantly better than other baselines, with improvements of 5.6%, 8.2%, and 15.8% over the best baseline in BLEU4, METEOR, and CIDEr, respectively.

With the advent of the Internet of Things (IoT) era and 5G, ubiquitous sensing devices (e.g., smartphones, surveillance sites, and security cameras) have been widely used in various fields, resulting in the generation of a huge amount of monitoring data. The rise of federated learning makes it possible to leverage monitoring data to train deep neural networks through cloud-edge collaboration without compromising privacy. However, the non identically and independently distributed (called Non-IID) data collected by IoT devieces creates a client drift phenomenon, resulting in a slow convergence of the global model. To this end, we propose a new Federated learning framework based on Gradient Variance Reduction with a correction weight control mechanism and Global gradient descent with Momentum, named FedGVRGM to conduct gradient correction and reduce the negative impacts of prediction parameters. Specifically, in the local training phase, FedGVRGM combines gradient variance reduction with a correction weight control mechanism to further correct the local model parameters, thus reducing the dispersion of model parameters among clients. In the global aggregation phase, FedGVRGM integrates the historical change states of the global model through the gradient descent with momentum to reduce the oscillations and improve the convergence speed of the global model. We refer to the above methods of gradient adjustment in the local and global training phases as FedGVR and FedGM, respectively. Numerous evaluations are conducted on CIFAR-100, CIFAR-10, and MNIST datasets to prove that FedGVRGM has a faster convergence rate than other stateof-the-art approaches such as Federated Averaging (FedAvg), FedProx, FedReg, FedGVR, and FedGM.

Abstract In edge computing, though distributed training of Deep Neural Networks (DNNs) is expected to exchange massive gradients between parameter servers and working nodes, the high communication cost constrains the training speed. To break this limitation, gradient compression algorithms expect the ultimate compression ratio at the expense of the accuracy of the trained model. Therefore, new gradient compression techniques are necessary to ensure both communication efficiency and model accuracy. This paper introduces a novel technique—an Adaptive Sparse Ternary Gradient Compression (ASTC) scheme, which relies on the number of gradients in model layers to compress gradients. ASTC establishes the model compression selection criterion by gradients’ amount, compresses the network layer that meets the model’s standard, evaluates the gradients’ importance based on entropy to adaptively perform sparse compression, and finally conducts ternary quantization compression and a lossless code scheme on sparse gradients. Using public datasets (MNIST, CIFAR-10, Tiny ImageNet) and deep learning models (CNN, LeNet5, ResNet18) for experimental evaluation, we exhibit excellent results that the training efficiency of ASTC is about 1.6 times, 1.37 times, and 1.1 times higher than that of Top-1, AdaComp, and SBC, respectively. Furthermore, ASTC can be improved by an average of about $$1.9\%$$ 1.9% in training accuracy compared with the above approaches.

In edge networks, distributed computing resources have been widely utilized to collaboratively perform a machine learning task by multiple nodes. However, the model training time in heterogeneous edge networks is becoming longer because of excessive computation and delay caused by slow nodes, namely, stragglers. The parameter server even abandons stragglers which fail to return the outcome within a reasonable deadline, called straggler dropout, decreasing the model accuracy. To optimize the computation cost and maintain the model accuracy, we focus on mitigating the heavy computation of stragglers and preventing straggler dropout. Therefore, we propose a novel scheme named dynamic grouping and heterogeneity-aware gradient coding (DGH-GC) to tolerate stragglers by employing dynamic grouping and gradient coding. DGH-GC evenly distributes stragglers in each group and encodes gradients based on their computation capacity to prevent them drop out. However, DGH-GC exacerbates the communication burden by making data duplication to tolerate stragglers. Relying on the scheme, we further propose an algorithm called DGH-(GC)2 to compress transferred gradients in both upstream communication and downstream communication. Experimental evaluations prove that DGH-(GC) outperforms all state-of-the-art methods and DGH-(GC)2 further speeds up the convergence time of the trained model and saves about 26% average iteration time compared to the DGH-(GC).

In the transportation industry, task offloading services of edge-enabled Internet of Vehicles (IoV) are expected to provide vehicles with the better Quality of Experience (QoE). However, the various status of diverse edge servers and vehicles, as well as varying vehicular offloading modes, make a challenge of task offloading service. Therefore, to enhance the satisfaction of QoE, we first introduce a novel QoE model. Specifically, the emerging QoE model restricted by the energy consumption: 1) intelligent vehicles equipped with caching spaces and computing units may work as carriers; 2) various computational and caching capacities of edge servers can empower the offloading; and 3) unpredictable routings of the vehicles and edge servers can lead to diverse information transmission. We then propose an improved deep reinforcement learning (DRL) algorithm named PS-DDPG with the prioritized experience replay (PER) and the stochastic weight averaging (SWA) mechanisms based on deep deterministic policy gradients (DDPG) to seek an optimal offloading mode, saving energy consumption. Specifically, the PER scheme is proposed to enhance the availability of the experience replay buffer, thus accelerating the training. Moreover, reducing the noise in the training process and thus stabilizing the rewards, the SWA scheme is introduced to average weights. Extensive experiments certify the better performance, i.e., stability and convergence, of our PS-DDPG algorithm compared to existing work. Moreover, the experiments indicate that the QoE value can be improved by the proposed algorithm.

In the transportation industry, task offloading services of edge intelligent Internet of Vehicles (IoV) are expected to provide vehicles with the better Quality of Experience (QoE). However, the various status of diverse edge servers and vehicles, as well as varying vehicular offloading modes, make a challenge of task offloading service. Therefore, to enhance the satisfaction of QoE, we first introduce a novel QoE model. Specifically, the emerging QoE model restricted by the energy consumption, (1) intelligent vehicles equipped with caching spaces and computing units may work as carriers; (2) various computational and caching capacities of edge servers can empower the offloading; (3) unpredictable routings of the vehicles and edge servers can lead to diverse information transmission. We then propose an improved deep reinforcement learning (DRL) algorithm named RA-DDPG with the prioritized experience replay (PER) and the stochastic weight averaging (SWA) mechanisms based on deep deterministic policy gradients (DDPG) to seek an optimal offloading mode, saving energy consumption. Extensive experiments certify the better performance, i.e., stability and convergence, of our RA-DDPG algorithm compared to existing work. Moreover, the experiments indicate that the QoE value can be improved by the proposed algorithm.

Due to the low quality of the existing online graph partition algorithm, the graph partition problem is solved through the Cat Swarm Optimization (CSO) algorithm to improve the partition quality. To avoid falling into the local optimum with CSO, an improved graph partition approach based on Cat Swarm Optimization with the Cauchy mutation and the Inertia weight (CICSO) was proposed. CICSO adopts the Cauchy mutation to update the optimal position, which can increase the accuracy of graph partition. Meanwhile, the self-adaptive inertia weight with the dynamic change is introduced in the tracking mode to increase the convergence speed and stability. Experimental results show that the improved cat algorithm CICSO has better performance than the standard cat algorithm in terms of partition quality and convergence time, compared with the LDG, FENNEL, and the standard CSO.

The real-time time series data of multiple water quality parameters are obtained from the water sensor networks deployed in the water supply network. The accurate and efficient detection and warning of contamination events to prevent pollution from spreading is one of the most important issues when the pollution occurs. In order to comprehensively reduce the event detection deviation, a Temporal Abnormal Event Detection Algorithm for Multivariate time series data (M-TAEDA) was proposed. In M-TAEDA, first, Back Propagation neural network models are adopted to analyze the time series data of multiple water quality parameters and calculate the possible outliers. Then, M-TAEDA algorithm determines the potential contamination events through Bayesian sequential analysis to estimate the probability of a contamination event. Finally, it can make decision based on the multiple event probabilities fusion in the water supply system. The experimental results indicate that the proposed M-TAEDA algorithm can obtain the 90% accuracy with BP neural network model and improve the rate of detection about 40% and reduce the false alarm rate about 45%, compared with the temporal event detection of Single Variate Temporal Abnormal Event Detection Algorithm (S-TAEDA).

Wireless sensor networks have a wide range of potential, practical and useful applications. However, there are many challenging problems that need to be addressed for efficient operation of wireless sensor networks in real applications. One of the fundamental and important problems in sensor networks is the topology control problem since most sensors are equipped with non-rechargeable batteries and the density of deployed sensors is very high. Topology control needs to reduce the power-consumption and extend the lifetime of sensor networks while satisfying certain application requirements. To energy-efficient control the topology structure of sensor networks, the common approaches are to adjust the transmission power of sensors and to dynamically schedule sensor's cycles. In this paper, we survey the state-of-the art topology control techniques, present an overview and analysis of the solutions proposed in recent research literature.