State-of-the-art (SOTA) fact-checking systems combat misinformation by employing autonomous LLM-based agents to decompose complex claims into smaller sub-claims, verify each sub-claim individually, and aggregate the partial results to produce verdicts with justifications (explanations for the verdicts). The security of these systems is crucial, as compromised fact-checkers can amplify misinformation, but remains largely underexplored. To bridge this gap, this work introduces a novel threat model against such fact-checking systems and presents Fact2Fiction, the first poisoning attack framework targeting SOTA agentic fact-checking systems. Fact2Fiction employs LLMs to mimic the decomposition strategy and exploit system-generated justifications to craft tailored malicious evidences that compromise sub-claim verification. Extensive experiments demonstrate that Fact2Fiction achieves 8.9%-21.2% higher attack success rates than SOTA attacks across various poisoning budgets and exposes security weaknesses in existing fact-checking systems, highlighting the need for defensive countermeasures.

Publisher Copyright: © 2026 Owner/Author.

State-of-the-art single-agent claim verification methods struggle with complex claims that require nuanced analysis of multifaceted evidence. Inspired by real-world professional fact-checkers, we propose DebateCV, the first debate-driven claim verification framework powered by multiple LLM agents. In DebateCV, two Debaters argue opposing stances to surface subtle errors in single-agent assessments. A decisive Moderator is then required to weigh the evidential strength of conflicting arguments to deliver an accurate verdict. Yet, zero-shot Moderators are biased toward neutral judgments, and no datasets exist for training them. To bridge this gap, we propose Debate-SFT, a post-training framework that leverages synthetic data to enhance agents' ability to effectively adjudicate debates for claim verification. Results show that our methods surpass state-of-the-art non-debate approaches in both accuracy (across various evidence conditions) and justification quality.

Motivated by real-world applications, we study the problem of decentralized online learning with dynamic feedback delays in the presence of malicious data generators under different threat models. In this problem, multiple agents collaborate to classify the features of streaming data samples generated online and receive dynamically delayed feedback on the ground-truth labels. While some data generators are benign, others—due to internal motives or external factors such as cyberattacks—may maliciously manipulate data features to compromise the classification performance. In this work, we first investigate the targeted attacks by malicious data generators, i.e., feature manipulation with aims to gain preferred classification outcomes from the agents. In response, we propose two robust algorithms,RDOC-TOandRDOC-TC, countering ordinary and clairvoyant adversaries that can access certain outdated and the latest classification models of the agents, respectively. Subsequently, we address the untargeted attacks by malicious data generators, which aim to disrupt the classification outcomes without targeting any particular class, by proposing another algorithm,RDOC-U. Our theoretical analysis establishes that all three proposed algorithms achieve sublinear regret bounds. The evaluations conducted in the application of network traffic classification with two real-world datasets demonstrate the competitiveness of the proposed algorithms compared to advanced baselines.

Person detection in videos is vital for area admission and public safety. Existing studies have made significant progress in improving the accuracy of this task on the cloud. Meanwhile, with people's increasing awareness of privacy protection, there is a surging demand for privacy not being transmitted and processed by the cloud. Thus, providing services on edges becomes a promising solution. The dilemma is that edges are typically resource-constrained and cannot support the deployment of large models. However, tiny models that fit resource-constrained edges generally have unsatisfactory performance in accuracy and efficiency. To this end, we propose a Logical Correction Enabled Collaborative Person Detection Inference (LC-CPDI) framework for resource-constrained edges. First, we formulate the problem studied with a delay minimization objective. Second, we design a logical correction scheme to perceive abnormal predictions and perform corrections to improve accuracy. Third, a hybrid position prediction algorithm is proposed to replace time-consuming inference for simple scenarios. Finally, we design a collaborative inference scheme that enables frame outsourcing to idle edges to reduce the inference delay. We implemented LC-CPDI on a testbed designed with commercial edges. The experiments on real-world datasets show the effectiveness of LC-CPDI with up to 41.8% delay reduction on average and near 2% recall improvement.

In recent years, chaos-based multimedia cryptosystems have gained prominence due to their nonlinear properties, such as sensitivity to initial conditions and long-term unpredictability, which parallel the cryptographic requirements of confusion and diffusion. However, many such systems lack standardization and deviate from secure design principles, resulting in practical vulnerabilities. This article presents a comprehensive cryptanalysis of a widely cited video cryptosystem. The target system combines S-box substitution—derived from either a 12D chaotic map or the Ikeda delay differential equation (DDE)—with a Cipher Block Chaining (CBC) diffusion scheme. Through an analysis of the encryption structure, fundamental design defects are identified. Specifically, the CBC diffusion mechanism lacks key dependence, and the system exhibits an excessive reliance on fixed, publicly exposed S-boxes. These vulnerabilities are demonstrated through chosen-plaintext attacks, known-plaintext attacks, and chosen-ciphertext attacks. Experimental results confirm that the S-box can be fully reconstructed, thereby facilitating the complete decryption of video content in the absence of the secret key. Beyond exposing vulnerabilities, this study offers constructive contributions by proposing six concrete improvement strategies, including dynamic key generation, structural enhancement, and dynamic S-box indexing. This work provides practical suggestions for designing secure chaos-based multimedia cryptosystems, offering valuable references for future research and promoting the development of efficient encryption systems.

Ride-sharing represents a sustainable transportation paradigm that is beneficial to human, society, and environment. Common ride-sharing pricing approaches determine prices for riders through optimizing one or more figures of merit, e.g., the profit or revenue. However, they overlook an important issue—fairness—which, when perceived by the riders, can critically affect their degree of satisfaction. In this work, we take the initiative to consider an intuitive and appropriate notion of individual fairness called fairness-in-hindsight for riders in ride-sharing pricing. We study the problem of online fair pricing of shared rides (which allow multiple riders to share one ride) with an aim to maximize the profit of the ride-sharing operator/platform. We design an online fair ride-sharing pricing algorithm called OnFairRP, which comprises phases oflearning, transition, and exploitation. We prove that OnFairRP has a sub-linear regret bound and can guarantee the fairness between riders. Our extensive performance evaluations using real-world data traces of ride-sharing demonstrate the advantages of OnFairRP over benchmarking schemes including commonly used methods with or without fairness guarantee.

To support delay-critical applications, migrating services from the cloud to edge servers (ESs) can effectively reduce service delay. However, in such a resource-constrained scenario, providing heterogeneous services to meet diverse user needs poses significant challenges. Specifically, ESs have limited computational resources and are often energy-constrained, complicating service placement and provisioning. Existing studies propose collaboration schemes to improve resource utilization and reduce service delay. Nevertheless, these works heavily depend on full-service coverage by the cloud or rely on coarse-grained (e.g., time-cycle level) strategies that inevitably waste energy in idle time slots, which are unsuitable for energy-constrained settings and dynamic edge environments. To address these challenges, we propose a novel edge-edge collaboration method tailored for heterogeneous edge service provisioning in energy-constrained networks. First, we formulate the heterogeneous service provisioning problem with the objective of delay minimization under energy constraints and prove its NP-hardness. Our framework leverages latest task statistics to decide the service lifetime in a fine-grained time slot level, so that enables edge collaboration to maximize resource utilization and adaptability in resource-limited conditions. Specifically, we decompose the problem into three subproblems: service placement, service lifetime decision, and task scheduling, and we design targeted lightweight solutions for each, ensuring low delay and efficient energy usage. Finally, we validate our method through comprehensive simulations on real-world datasets and implement it on a testbed with three ESs. Results demonstrate that our approach reduces service delays by an average of 69.6% across various energy-constrained scenarios.

Person re-identification (ReID) in distributed surveillance networks presents significant networking challenges, particularly in coordinating multiple edge devices for real-time processing. While cloud-based solutions offer powerful computational capabilities, they introduce substantial network latency, hindering real-time performance in multi-camera indoor environments. To address these challenges, we present a distributed edge computing architecture that enables real-time person reidentification. Our system introduces two key networking innovations: (1) an$N$-frame feature matching mechanism that enhances identification accuracy at network edges, and (2) a dynamic load balancing framework that efficiently distributes processing tasks across edge nodes when network congestion occurs. As a foundation for this research, we introduce BNBUMTMC, a mixed dataset comprising both images and videos from multiple cameras, specifically tailored for indoor MTMCReID scenarios. Through extensive deployment in a university building environment with 10 cameras and multiple edge devices, we demonstrate that our system achieves high identification accuracy while significantly reducing network transmission latency compared to cloud-based approaches. Our experience provides practical insights into designing and implementing distributed edge computing systems for real-time surveillance applications.

The integration of edge computing with deep neural networks (DNNs) is crucial for intelligent industrial cyber-physical systems. Typically, deploying DNNs on heterogeneous edge devices relies on methods like model compression and partitioning. However, these approaches often result in homogeneous models across devices. This homogeneity limits the collective capability of edge computing systems, particularly in terms of generalization to diverse data distributions and adaptation to dynamic industrial environments. In this work, we propose to treat each DNN on an edge device as an independent model, aggregating their capabilities via ensemble learning to enhance generalization and dynamic adaptability. To realize this, we introduce the Adaptive Ensemble Knowledge Distillation Framework (AEKDF), combining cloud-based model training with edge computing based collaborative inference. In the cloud, AEKDF develops an enhanced Born Again Network that generates diverse, lightweight models tailored to specific edge devices through knowledge distillation. This process ensures model diversity which is critical to effective ensemble learning. On the edge, AEKDF employs an adaptive ensemble technique that aggregates prediction logits across devices, enabling rapid adaptation to changing environments and maintaining inference efficiency. Our extensive evaluations conducted on a realistic prototype demonstrate the substantial boost in predictive performance achieved by our AEKDF, showcasing a 4% to 10% accuracy improvement on the CIFAR-100 compared to conventional single-model approaches, while maintaining low latency.

Decentralized online convex optimization (DOCO), as a pivotal computational paradigm in machine learning, has been applied to many critical tasks. However, existing DOCO algorithms, due to their excessive emphasis on the worst-case theoretical performance, appear to be overly cautious in making decisions across all possible cases, especially in real-world applications where the worst cases actually hardly occur. Therefore, these existing approaches typically are limited in performance in practice. To avoid such pessimistic strategies, we propose to study the approach of augmenting DOCO with machine-learned predictions that can guide the decision-making process. We present an overview of the problem along with the preliminary results and outlook in this work.

Decentralized online optimization, a pivotal paradigm in machine learning, involves multiple agents making online decisions cooperatively in a decentralized network. Despite its outstanding capabilities in processing large-scale streaming data, the ubiquitous existence of malicious agents, capable of disseminating arbitrary information among their neighbors and undetectable a priori, poses a severe threat to the reliability and efficacy of existing decentralized online optimization solutions. In response to the above critical vulnerability in practice, we take the first step to properly address the threat posed by malicious agents. We propose ROOO, a novel robust decentralized online optimization algorithm, specifically designed to counteract the detrimental impact of malicious agents. Our theoretical analysis shows that the regret bound of ROOO is sub-linear, indicating that, over time, its performance progressively approximates that of an offline oracle operating with the benefit of hindsight. Empirical evaluations in two networking applications, including opportunistic channel selection and mobile crowdsensing, further validate our theoretical results and demonstrate the competitiveness of ROOO compared to several advanced baselines.

Collaborative edge computing has been widely advo-cated by network operators and service providers to promote the quality of service (QoS), provisioning diverse delay-sensitive and computation-intensive applications. Existing studies mainly focus on cloud-edge collaboration, since cloud servers have massive resources to provide diverse services and edge servers can provide low-delay services with close proximity to end users. However, in scenarios that capture privacy, e.g., personal bioinformation and business areas, there is a great need for zero cloud involvement. Moreover, current edge servers are typically energy-constrained, which poses great challenges in enabling high-QoS services in ever-densely deployed edge networks. To tackle these issues, in this paper, we study the energy-constrained edge-edge collaboration problem. First, we formulate the edge-edge collaboration with delay minimization and energy reduction aims and prove its NP-hardness. Second, we propose a novel Fine-Grained Service Lifetime Optimization (FGSLO) scheme as a possible solution. The problem is then transformed and decoupled into three sub-problems, namely service placement, service lifetime decision, and task scheduling, which are solved by our proposed method, respectively. Finally, real-world data-driven experimental results show that FGSLO is capable of reducing 21.4%~90.1 % system delay in different energy-constrained scenarios, compared to baselines without service lifetime control.

The online linear optimization paradigm is important to many real-world network applications as well as theoretical algorithmic studies. Recent studies have made attempts to augment online linear optimization with machine-learned predictions of the cost function that are meant to improve the performance of the algorithms. However, they fail to address the critical case in practical systems where the predictions can be arbitrarily bad. In this work, we take the first step to study the problem of online linear optimization with a dynamic number of arbitrarily bad machine-learned predictions per round and propose an algorithm termed OLOAP. Our theoretical analysis shows that, when the qualities of the predictions are satisfactory, OLOAP achieves a regret bound of O(logT), which circumvents the tight lower bound of Ω($\sqrt T $) for the vanilla problem of online linear optimization (i.e., the one without any predictions). Meanwhile, the regret of our algorithm is never worse than O($\sqrt T $) irrespective of the qualities of predictions. In addition, we further derive a lower bound for the regret of the studied problem, which demonstrates that OLOAP is near-optimal. We consider two important network applications and conduct extensive evaluations. Our results validate the superiority of our algorithm over state-of-the-art approaches.

Collaborative edge computing enables low service delay for many delay-sensitive Internet of Things applications through edge-edge and edge-cloud collaborations. Due to the limited edge resources and varying task demands, optimizing Joint Service Placement and Task Offloading (JSPTO) becomes crucial in minimizing overall processing delays. However, existing JSPTO methods overlook the impact of service startup delay, which may undermine total latency reduction, especially in scenarios with large startup delays. This paper introduces an online JSPTO method that integrates the consideration of service startup delay to enhance task offloading efficiency. However, a significant challenge is ensuring timely service response with large startup delays. We formulate this problem as an integer linear programming problem, aiming to minimize the total service startup and task processing delay. We propose a novel algorithm called SD-JSPTO, which performs online JSPTO in the presence of large startup delays. Theoretical performance analyses reveal that SD-JSPTO attains a near-optimal solution within polynomial time, demonstrating a competitive ratio of $1 + \frac{{{A_2}}}{{V{T^{{\text{opt}}}}}}$. Experimental evaluations demonstrate that our method significantly reduces the total delay by no less than 18.72% compared to state-of-the-art baseline methods while preserving system stability.

The prevalence of fake news across various online sources has had a significant influence on the public. Existing Chinese fake news detection datasets are limited to news sourced solely from Weibo. However, fake news originating from multiple sources exhibits diversity in various aspects, including its content and social context. Methods trained on purely one single news source can hardly be applicable to real-world scenarios. Our pilot experiment demonstrates that the F1 score of the state-of-the-art method that learns from a large Chinese fake news detection dataset, Weibo-21, drops significantly from 0.943 to 0.470 when the test data is changed to multi-source news data, failing to identify more than one-third of the multi-source fake news. To address this limitation, we constructed the first multi-source benchmark dataset for Chinese fake news detection, termed MCFEND, which is composed of news we collected from diverse sources such as social platforms, messaging apps, and traditional online news outlets. Notably, such news has been fact-checked by 14 authoritative fact-checking agencies worldwide. In addition, various existing Chinese fake news detection methods are thoroughly evaluated on our proposed dataset in cross-source, multi-source, and unseen source ways. MCFEND, as a benchmark dataset, aims to advance Chinese fake news detection approaches in real-world scenarios.

Recent analyses have disclosed that existing rumor detection techniques, despite playing a pivotal role in countering the dissemination of misinformation on social media, are vulnerable to both white-box and surrogate-based black-box adversarial attacks. However, such attacks depend heavily on unrealistic assumptions, e.g., modifiable user data and white-box access to the rumor detection models, or appropriate selections of surrogate models, which are impractical in the real world. Thus, existing analyses fail to uncover the robustness of rumor detectors in practice. In this work, we take a further step towards the investigation about the robustness of existing rumor detection solutions. Specifically, we focus on the state-of-the-art rumor detectors, which leverage graph neural network based models to predict whether a post is rumor based on the Message Propagation Tree (MPT), a conversation tree with the post as its root and the replies to the post as the descendants of the root. We propose a novel black-box attack method, HMIA-LLM, against these rumor detectors, which uses the Large Language Model to generate malicious messages and inject them into the targeted MPTs. Our extensive evaluation conducted across three rumor detection datasets, four target rumor detectors, and three baselines for comparison demonstrates the effectiveness of our proposed attack method in compromising the performance of the state-of-the-art rumor detectors.

Phone number recycling (PNR) refers to the event wherein a mobile operator collects a disconnected number and reassigns it to a new owner. It has posed a threat to the reliability of the existing authentication solution for e-commerce platforms. Specifically, a new owner of a reassigned number can access the application account with which the number is associated, and may perform fraudulent activities. Existing solutions that employ a reassigned number database from mobile operators are costly for e-commerce platforms with large-scale users. Thus, alternative solutions that depend on only the information of the applications are imperative. In this work, we study the problem of detecting accounts that have been compromised owing to the reassignment of phone numbers. Our analysis on Meituan’s real-world dataset shows that compromised accounts have unique statistical features and temporal patterns. Based on the observations, we propose a novel model called temporal pattern and statistical feature fusion model (TSF) to tackle the problem, which integrates a temporal pattern encoder and a statistical feature encoder to capture behavioral evolutionary interaction and significant operation features. Extensive experiments on the Meituan and IEEE-CIS datasets show that TSF significantly outperforms the baselines, demonstrating its effectiveness in detecting compromised accounts due to reassigned numbers.

Ride-hailing is a sustainable transportation paradigm where riders access door-to-door traveling services through a mobile phone application, which has attracted a colossal amount of usage. There are two major planning tasks in a ride-hailing system: 1) matching, i.e., assigning available vehicles to pick up the riders; and 2) repositioning, i.e., proactively relocating vehicles to certain locations to balance the supply and demand of ride-hailing services. Recently, many studies of ride-hailing planning that leverage machine learning techniques have emerged. In this article, we present a comprehensive overview on latest developments of machine learning-based ride-hailing planning. To offer a clear and structured review, we introduce a taxonomy into which we carefully fit the different categories of related works according to the types of their planning tasks and solution schemes, which include collective matching, distributed matching, collective repositioning, distributed repositioning, and joint matching and repositioning. We further shed light on many real-world data sets and simulators that are indispensable for empirical studies on machine learning-based ride-hailing planning strategies. At last, we propose several promising research directions for this rapidly growing research and practical field.

The multi-armed bandit (MAB) models have attracted significant research attention due to their applicability and effectiveness in various real-world scenarios such as resource allocation in uncertain environments, online advertising, and dynamic pricing. As an important branch, the adversarial multi-armed bandit problems with delayed feedback have been proposed and studied by many researchers recently where a conceptual adversary strategically selects the reward distributions associated with each arm to challenge the learning algorithm and the agent experiences a bunch of delays in receiving the corresponding reward feedback from different users after taking an action on them. However, the existing models restrict the feedback to being generated from only one user, which makes models inapplicable to the prevailing scenarios of multiple users (e.g. ad recommendation for a group of users). In this paper, we consider that the delayed feedback results are from multiple users and are unrestricted on internal distribution while the feedback delay is arbitrary and unknown to the player in advance. Also, for different users in a round, the delays in feedback have no assumption of latent correlation. Thus, we formulate an adversarial multi-armed bandit problem with multi-user delayed feedback and design a modified EXP3 algorithm named MUD-EXP3, which makes a decision at each round by considering the importance-weighted estimator of the received feedback from different users. On the premise of known terminal round index$T$, the number of users$M$, the number of arms$N$, and upper bound of delay$d_{max}$, we prove a regret of$\mathcal {O}(\sqrt{TM^{2}\ln {N}(N\mathrm{e}+4d_{max})})$. Furthermore, for the more common case of unknown$T$, an adaptive algorithm named AMUD-EXP3 is proposed with a sublinear regret concerning$T$. Finally, extensive experiments are conducted to indicate the correctness and effectiveness of our algorithms in dynamic environments.

Deep learning (DL) powered real-time applications usually need continuous training using data streams generated over time and across different geographical locations. Enabling data offloading among computation nodes through model training is promising to mitigate the problem that devices generating large datasets may have low computation capability. However, offloading can compromise model convergence and incur communication costs, which must be balanced with the long-term cost spent on computation and model synchronization. Therefore, this paper proposes EdgeC3, a novel framework that can optimize the frequency of model aggregation and dynamic offloading for continuously generated data streams, navigating the trade-off between long-term accuracy and cost. We first provide a new error bound to capture the impacts of data dynamics that are varying over time and heterogeneous across devices, as well as quantifying varied data heterogeneity between local models and the global one. Based on the bound, we design a two-timescale online optimization framework. We periodically learn the synchronization frequency to adapt with uncertain future offloading and network changes. In the finer timescale, we manage online offloading by extending Lyapunov optimization techniques to handle an unconventional setting, where our long-term global constraint can have abruptly changed aggregation frequencies that are decided in the longer timescale. Finally, we theoretically prove the convergence of EdgeC3 by integrating the coupled effects of our two-timescale decisions, and we demonstrate its advantage through extensive experiments performing distributed DL training for different domains.

Collaborative Mobile Edge Computing (MEC) has emerged as a promising approach to provide low service latency for computation-intensive Internet of Things applications, facilitated by the cooperation of edge-edge and edge-cloud resources. However, existing collaborative MEC methods typically restrict the collaborative processing between any two Edge Servers (ESs) or one ES and the cloud server for a task request, limiting the exploitation of available resources on other ESs. Moreover, these conventional methods rely on offline task partitioning, potentially leading to extended make-span, especially when ES computing capacities exhibit heterogeneity. In this paper, we propose an innovative method named SMCoEdge. This method performs dynamic parallel multi-ES selection and workload allocation in heterogeneous collaborative MEC environments, thus simultaneously enabling multiple ESs' idle resources to accelerate task processing. We formulate our problem into an online linear programming problem, with the objective of minimizing task computing and transmission make-spans. To enhance computational efficiency, we decompose the problem into two stages: multi-ES selection and workload allocation. Then, we propose an online Deep Reinforcement Learning based Simultaneous Multi-ES Offloading (DRL-SMO) algorithm along with a top-$k$deep Q-learning network model to effectively solve our problem, where an efficient algorithm is proposed to achieve the optimal solution for the workload allocation stage. Furthermore, we provide a theoretical performance analysis, demonstrating that the DRL-SMO algorithm achieves a near-optimal solution for our problem within an approximate linear time complexity. Finally, our extensive experimental results demonstrate the substantial advantages of our method. It consistently reduces the average make-span by 19.63% and keeps a lower offloading failure rate, when compared to state-of-the-art methods. These findings underline the efficacy of our method in enhancing collaborative MEC performance.

CDN is a crucial Internet infrastructure ensuring quick access to Internet content. With the expansion of CDN scenarios, beyond delay, resource types like bandwidth and CPU are also important for CDN performance. Our measurements highlight the distinct impacts of various resource types on different CDN requests. Unfortunately, mainstream CDN server selection schemes only consider a single resource type and are unable to choose the most suitable servers when faced with diverse resource types. To fill this gap, we propose Polygon, a QUIC-powered CDN server selection system that is aware of multiple resource demands. Being an advanced transport layer protocol, QUIC equips Polygon with customizable transport parameters to enable the seamless handling of resource requirements in requests. Its 0-RTT and connection migration mechanisms are also utilized to minimize delays in connection and forwarding. A set of collaborative measurement probes and dispatchers are designed to support Polygon, being responsible for capturing various resource information and forwarding requests to suitable CDN servers. Real-world evaluations on the Google Cloud Platform and extensive simulations demonstrate Polygon’s ability to enhance QoE and optimize resource utilization. The results show up to a 54.8% reduction in job completion time, and resource utilization improvements of 13% in bandwidth and 7% in CPU.

No abstract available.

To understand different aspects of online human behaviors, e.g., the public stances toward various social and political issues, contextual target-specific stance detection has become one of the most important studies on social media. Considering the lack of appropriate data for the studies of contextual target-specific stance detection on Twitter, which is one of the most popular online social platforms worldwide, we introduce CTSDT, a new dataset that consists of a large number of annotated target-specific conversations collected from Twitter. Furthermore, we propose a new contextual target-specific stance detection model called ConMulAttn, which is the first method that can learn both the contents of the posts and the concrete relationships between the posts in a conversation. We conduct extensive evaluation using CTSDT as well as another two popular datasets, CreateDebate and ConvinceMe, for contextual target-specific stance detection. The evaluation results validate the necessity of introducing our dataset CTSDT. Besides, according to the evaluation results, our proposed model ConMulAttn can outperform the state-of-the-art contextual target-specific stance detection method by up to 25% in F1score, indicating the effectiveness and superiority of our solution. Our study has the potential to assist policymakers in utilizing conversation data from online social platforms to efficiently gain real-time insights into public stances on target topics, such as vaccination.

To meet the growing need of mobile traffic with Ultra-Reliable and Low Latency Communication (URLLC) requirement, 5G New Radio (NR) is extending from licensed band to unlicensed band on which Wi-Fi has already been operated, resulting in coexisting NR/Wi-Fi network. Existing works have made great efforts on throughput and latency of coexisting NR/Wi-Fi network. However, excessive NR requests offloaded from licensed band lead to unfair utilization of unlicensed band, which further causes unsatisfaction on URLLC and performance degradation of Wi-Fi. In this paper, we propose a novel Reinforcement Learning based Transmission Revoking Approach (RL-TRA) to address this problem aiming at fairer utilization of unlicensed band restrained by URLLC. Firstly, we formulate the coexistence problem of NR/Wi-Fi as integer non-linear programming and show its NP-hardness. Secondly, we decompose the problem into three sub-problems, namely redundancy determining, request scheduling, and transmission revoking. The former two sub-problems are solved with our proposed method to satisfy URLLC requirement. We further propose a novel transmission revoking mechanism when tackling transmission revoking sub-problem, aiming at maintaining fairness of coexisting NR/Wi-Fi network. Finally, simulation results verify the effectiveness of RL-TRA. By using our method, the fairness is improved by 16.5% averagely with only 1.77% loss on success rate of URLLC requests compared with baselines.

The integration of edge computing and deep neural networks (DNNs) holds great promise for enhancing application intelligence. Edge devices generate or collect vast amounts of data, which DNNs can leverage to make informed decisions. Nevertheless, the limited resources of edge devices pose a significant challenge for deploying DNNs. To accommodate some edge devices (e.g. smart watches), lightweight models are often required. However, the accuracy of these models may not meet user expectations. In this paper, we present EKDF, an ensemble knowledge distillation framework that crafts lightweight models for collaborative DNN inferences. More specifically, we utilize knowledge distillation to compress DNN models. On this basis, we introduce multi-teacher joint supervision and dropout in knowledge distillation to improve model performance and preserve the diversity between the generated DNN models. This process produces a range of compact models of varying computational complexity for different edge devices. The experimental results demonstrate that our proposed EKDF can greatly improve the overall predictive ability.

Motivated by the real-world application of traffic classification at the network edge, we study the problem of robust decentralized online learning against malicious data generators that can manipulate their data features with an aim to gain preferred classification outcomes. Multiple agents cooperatively learn classification models to make online decisions. They periodically exchange their models, e.g., traffic classification models, between neighbors in a decentralized network and update local model parameters on the fly based on the models they have access to and feedback on the observed local data samples that are dynamically delayed. In this work, we propose two decentralized online learning algorithms, RDOC-O and RDOC-C, respectively against ordinary malicious and clairvoyant malicious data generators. Our theoretical performance analysis shows that the two algorithms have provable sub-linear individual regret bounds under mild conditions. To validate our analysis, extensive performance evaluations are conducted in the application of network traffic classification using two real-world data traces. Our results show that the two proposed algorithms compare favorably with an optimal offline classification model in the presence of malicious data generators, and they can achieve a steady-state F1score of around 0.85, which validates their effectiveness and makes them appealing in practice.

Deep learning (DL) powered real-time applications usually need continuous training using data streams generated geographically. Enabling data offloading among computation nodes through model training is promising to mitigate the problem that devices generating large datasets may have low computation capability. However, offloading can compromise model convergence and incur communication costs, which must be balanced with the cost spent on computation and model synchronization. Therefore, this paper proposes EdgeC3, a novel framework that can optimize the frequency of model aggregation and dynamic offloading for continuously generated data streams, navigating the trade-off between long-term accuracy and cost. We first provide a new error bound to capture the impacts of data dynamics that are varying over time and heterogeneous across devices. Based on the bound, we design a two-timescale online optimization framework. We periodically learn the synchronization frequency to adapt with uncertain future offloading and network changes. In the finer timescale, we manage online offloading by extending Lyapunov optimization techniques to handle an unconventional setting, where our long-term global constraint can have abruptly changed aggregation frequencies that are decided in the longer timescale. Finally, we theoretically prove the convergence of EdgeC3 by integrating the coupled effects of our two-timescale decisions, and we demonstrate its advantage through extensive experiments.

Target-specific stance detection on social media, which aims at classifying a textual data instance such as a post or a comment into a stance class of a target issue, is an emerging opinion mining paradigm of importance. An example application would be to overcome vaccine hesitancy in combating the coronavirus pandemic. Existing stance detection strategies rely merely on the individual instances which cannot always capture the expressed stance of a given target. We address a new task called conversational stance detection (CSD) which is to infer the stance toward a given target (e.g., COVID-19 vaccination) when given a data instance and its corresponding conversation thread. To carry out the task, we first propose a benchmarking CSD dataset with annotations of stances and the structures of conversation threads among the instances, which is based on six major social media platforms in Hong Kong. To infer the desired stances from both data instances and conversation threads, we propose a model called Branch-bidirectional encoder representations from transformers (BERT) that incorporates contextual information in conversation threads. Extensive experiments on our CSD dataset show that our proposed model outperforms all the baseline models that do not make use of contextual information. Specifically, it improves the F1 score by 10.3% compared with the state-of-the-art method in the SemEval-2016 Task 6 competition. This shows the potential of incorporating rich contextual information on detecting target-specific stances on social media platforms and suggests a more practical way to construct future stance detection tasks.

Online-to-Offline (O2O) e-commerce service platforms and their users are faced with various fraud risks. Among them, financial identity theft is a widely existing challenge. However, existing methods are insufficient to detect this type of fraud. In this paper, we address the financial identity theft detection problem in e-commerce services by leveraging access environment and behavior sequence. To explore the fraud patterns, we first make a detailed analysis using real cases of identity theft from Meituan, a leading O2O e-commerce platform in China. Our findings are twofold. First, fraudulent accounts sharing the same personal ID would have different access environments, such as devices and IP addresses. Second, a group of fraudulent accounts may have aggregations of devices, IP addresses, and delivery addresses. Based on these observations, we propose a hybrid method termed EnvIT to detect financial identity theft based on the heterogeneous graph and the behavior sequence. EnvIT is able to characterize the access environment and the historical behavior of the accounts. Furthermore, an attentive module is adopted to assign weights to different features automatically. We further evaluate EnvIT via extensive experiments using a real-world dataset from Meituan. Our experimental results demonstrate that EnvIt outperforms several baseline methods in fraudulent account detection and achieves an AUC of 0.9210.

Malicious data manipulation reduces the effectiveness of machine learning techniques, which rely on accurate knowledge of the input data. Motivated by real-world applications in network flow classification, we address the problem of robust online learning with delayed feedback in the presence of malicious data generators that attempt to gain favorable classification outcome by manipulating the data features. We propose online algorithms termed ROLC-NC and ROLC-C when the malicious data generators are non-clairvoyant and clairvoyant, respectively. We derive regret bounds for both algorithms and show that they are sub-linear under mild conditions. We further evaluate the proposed algorithms in network flow classification via extensive experiments using real-world data traces. Our experimental results demonstrate that both algorithms can approach the performance of an optimal static offline classifier that is not under attack, while outperforming the same offline classifier when tested with a mixture of normal and manipulated data.

Malicious data manipulation reduces the effectiveness of machine learning techniques, which rely on accurate knowledge of the input data. Motivated by real-world applications in network flow classification, we address the problem of robust online learning with delayed feedback in the presence of malicious data generators that attempt to gain favorable classification outcome by manipulating the data features. When the feedback delay is static, we propose online algorithms termed ROLC-NC and ROLC-C when the malicious data generators are non-clairvoyant and clairvoyant, respectively. We then consider the dynamic delay case, for which we propose online algorithms termed ROLC-NC-D and ROLC-C-D when the malicious data generators are non-clairvoyant and clairvoyant, respectively. We derive regret bounds for these four algorithms and show that they are sub-linear under mild conditions. We further evaluate the proposed algorithms in network flow classification via extensive experiments using real-world data traces. Our experimental results demonstrate that the proposed algorithms can approach the performance of an optimal static offline classifier that is not under attack, while outperforming the same offline classifier when tested with a mixture of normal and manipulated data.

To improve the application-level data efficiency, the scheduling of coflows, defined as a collection of parallel flows sharing the same objective, is prevailing in recent data centers. Meanwhile, optical circuit switches (OCS) are gradually applied to provide high data rate with low power consumption. However, so far few research outputs have covered the flow, let alone the coflow, scheduling in the context of OCS. In this work, we investigate coflow scheduling in OCS-based data centers. We first derive a novel operation called regularization processed respectively on the flow traffic demands and the flow start times, which can be efficiently implemented and reduce the circuit reconfiguration frequency dramatically. We then propose a 2-approximation algorithm, called Reco-Sin, for single coflow scheduling to minimize the coflow completion time (CCT). For multiple coflows, we derive Reco-Mul to minimize the total weighted CCT, which can transform any non-preemptive multi-coflow scheduling in packet switches to a scheduling scheme in OCS. Reco-Mul can achieve a constant approximation under the assumption that no tiny flows will be transmitted in OCS. To get rid of this assumption, we present another multiple coflow scheduling scheme, named Reco-Mul+, which has an approximation ratio of O(K). Here, K is the total number of coflows. Extensive simulations based on Facebook data traces show that our approaches outperform state-of-the-art schemes significantly, i.e., one single coflow can be finished up to 1.97× faster with Reco-Sin, and multiple coflows can be completed up to more than 2× faster with Reco-Mul and Reco-Mul+.

Network flow classification is essential to proper provisioning of Quality of Service (QoS). Conventional machine-learning based flow classification methods assume reliable knowledge of the flow features. However, in practice, malicious flow generators can manipulate the flow features to increase the likelihood of certain learning outcomes, e.g., in terms of the QoS requirement label. Training a classifier that is robust to such feature manipulation is imperative. In this work, we present a study on robust flow classification against malicious feature manipulation. We leverage a detailed system model to capture the relation between the classifier and malicious flow generators and propose a Stackelberggame based solution framework to train a robust classifier. We conduct extensive experimentation using real-world traces. For flows with manipulated features, the Stackelberg classifier trained by our solution framework significantly outperforms a non-robust classifier that is oblivious to manipulation, achieving accuracy close to that of the non-robust classifier on unmanipulated flows. Furthermore, the Stackelberg classifier on manipulated test flows is no worse than the non-robust classifier on unmanipulated flows.

To improve the application-level data efficiency, the scheduling of coflows, defined as a collection of parallel flows sharing the same objective, is prevailing in recent data centers. Meanwhile, optical circuit switches (OCS) are gradually applied to provide high data rate with low power consumption. However, so far few research outputs have covered the flow scheduling in the context of OCS, let alone the coflow scheduling problems. In this paper, we investigate coflow scheduling in the OCS-based data centers. We first derive a novel operation called regularization processed respectively on the flow traffic demands and the flow start times. Regularization can be efficiently implemented and reduce the circuit reconfiguration frequency dramatically. We then propose a 2-approximation algorithm, called Reco-Sin, for single coflow scheduling to minimize the coflow completion time (CCT). For multiple coflows, we derive another approximation algorithm, called Reco-Mul, to minimize the total weighted CCT, which can transform any non-preemptive multi-coflow scheduling in packet switches to that in OCS. Extensive simulations based on Facebook data traces show that Reco-Sin and Reco-Mul outperform state-of-the-art schemes significantly, i.e., one single coflow can be finished up to 2.72× faster with Reco-Sin, and multiple coflows can be completed up to 3.44× faster with Reco-Mul.

In edge-cloud computing, a set of servers (called edge servers) are deployed near the mobile devices to allow these devices to offload their jobs to and subsequently obtain their results from the edge servers with low latency. One fundamental problem in edge-cloud systems is how to dispatch and schedule the jobs so that the job response time (defined as the interval between the release of the job and the arrival of the computation result at the device) is minimized. In this paper, we propose a general model for this problem, where the jobs are generated in arbitrary order and at arbitrary times at the mobile devices and then offloaded to servers with both upload and download delays. Our goal is to minimize the total weighted response time of all the jobs. The weight is set based on how latency-sensitive the job is. We derive the first online job dispatching and scheduling algorithm in edge-clouds, called OnDisc, which is scalable in the speed augmentation model; that is, OnDisc is (1 + ε)-speed O(1/ε)-competitive for any small constant ε > 0. Moreover, OnDisc can be easily implemented in distributed systems. We also extend OnDisc with a fairness knob to incorporate the trade-off between the average job response time and the degree of fairness among jobs. Extensive simulations based on a real-world data-trace from Google show that OnDisc can reduce the total weighted response time dramatically compared with heuristic algorithms.

Efficient virtual machine (VM) management can dramatically reduce energy consumption in data centers. Existing VM management algorithms fall into two categories based on whether the VMs’ resource demands are assumed to be static or dynamic. The former category fails to maximize the resource utilization as they cannot adapt to the dynamic nature of VMs’ resource demands. Most approaches in the latter category are heuristic and lack theoretical performance guarantees. In this paper, we formulate the dynamic VM management as a large-scale Markov decision process (MDP) problem and derive an optimal solution. Our analysis of real-world data traces supports our choice of the modeling approach. However, solving the large-scale MDP problem suffers from the curse of dimensionality. Therefore, we further exploit the special structure of the problem and propose an approximate MDP-based dynamic VM management method, called MadVM. We prove the convergence of MadVM and analyze the bound of its approximation error. Moreover, we show that MadVM can be implemented in a distributed system with at most two times of the optimal migration cost. Extensive simulations based on two real-world workload traces show that MadVM achieves significant performance gains over two existing baseline approaches in power consumption, resource shortage, and the number of VM migrations. Specifically, the more intensely the resource demands fluctuate, the more MadVM outperforms.

A coflow is a collection of related parallel flows that occur typically between two stages of a multi-stage computing task in a network, such as shuffle flows in MapReduce. The coflow abstraction allows applications to convey their semantics to the network so that application-level requirements can be better satisfied. In this paper, we study the routing and scheduling of multiple coflows to minimize the total weighted coflow completion time (CCT). We first propose a rounding-based randomized approximation algorithm, called OneCoflow, for single coflow routing and scheduling. The multiple coflow problem is more challenging as coexisting coflows will compete for the same network resources, such as link bandwidth. To minimize the total weighted CCT, we derive an online multiple coflow routing and scheduling algorithm, called OMCoflow. We then derive a competitive ratio bound of our problem and prove that the competitive ratio of OMCoflow is nearly tight. To the best of our knowledge, this is the first online algorithm with theoretical performance guarantees which considers routing and scheduling simultaneously for multi-coflows. Compared with existing methods, OMCoflow runs more efficiently and avoids frequently rerouting the flows. Extensive simulations on a Facebook data trace show that OMCoflow outperforms the state-of-the-art heuristic schemes significantly (e.g., reducing the total weighted CCT by up to 41.8% and the execution time by up to 99.2% against RAPIER).

Motivated by practical scenarios, we study congestion games with failures. We investigate two models. The first model is congestion games with both resource and agent failures, where each agent chooses the same number of resources with the minimum expected cost. We prove that the game is potential and hence admits at least one pure-strategy Nash equilibrium (pure-NE). We also show that the Price of Anarchy and the Price of Stability are bounded (equal to 1 in some cases). The second model is congestion games with only resource failures (CG-CRF), where resources are provided in packages, and their failures can be correlated with each other. Each agent can choose multiple packages for reliability’s sake and utilize the survived one having the minimum cost. CG-CRF is shown to be not potential. We prove that it admits at least one pure-NE by constructing one efficiently. Finally, we discuss various applications of these two games in the networking field. To the best of our knowledge, this is the first paper studying congestion games with the coexistence of resource and agent failures, and we give also the first proof of the existence of a pure-NE in congestion games with correlated package failures.

A coflow is a collection of related parallel flows that occur typically between two stages of a multi-stage compute task in a network, such as shuffle flows in MapReduce. The coflow abstraction allows applications to convey their semantics to the network so that application-level requirements (e.g., minimizing the completion time of the slowest flow) can be better satisfied. In this paper, we study the routing and scheduling of multiple coflows to minimize the average coflow completion time (CCT). We first propose a rounding-based randomized approximation algorithm, called OneCoflow, for single coflow routing and scheduling. The multiple coflow problem is more challenging as coexisting coflows will compete for the same network resources such as link bandwidths. To minimize the average CCT, we derive an online multiple coflow routing and scheduling algorithm, called OMCoflow, and prove that it has a reasonably good competitive ratio. To the best of our knowledge, this is the first online algorithm with theoretical performance guarantees which considers routing and scheduling simultaneously for multi-coflows. Compared with existing methods, OMCoflow runs more efficiently, and it avoids the problem of frequently rerouting the flows. Extensive simulations on a Facebook data trace show that OMCoflow outperforms the state-of-the-art heuristic schemes significantly (e.g., reducing the average CCT by up to 41.8% and the execution time by up to 99.2% against RAPIER [28]).

Current large-scale computing services, such as online social networking and web searching, make the wired links in the data centers with an Ethernet infrastructure oversubscribed. Therefore, researchers consider to augment the data centers with wireless communication, called a hybrid data center network (HDCN), to improve the communication flexibility and network capacity. In this paper, we investigate how to use the wireless communication in hybrid DCNs from a cross-layer view. In the network layer, we propose a routing protocol to minimize the number of hops for data flows, and a congestion control protocol to reduce the congestion and deal with sporadic link failure. In the physical layer, we study the channel and power allocation problem with the SINR and QoS constraints in hybrid DCNs. In single channel scenarios, we prove the problem to be a geometric programming problem. In multi-channel scenarios, we prove the problem to be NP-hard and propose a Greedy based Online Channel and Power Allocation (GOCPA) algorithm. Our proposed protocols in network and physical layers collaborate to manage the wireless communication in hybrid DCNs. Extensive simulations show that our protocols can significantly increase the network throughput, decrease the latency, and moreover increase the robustness of the networks.

In Hybrid networks, which synergistically mix together wired and wireless links to achieve flexible and reliable communication, it is particularly challenging to routing selfish tasks since each task wish to finish transmission as early as possible and its decision could have impacts on the others. In this paper, we investigate the problem to route a given set of selfish tasks in hybrid networks. Under a unified cost model, the competitive behaviors of selfish players are modeled as a noncooperative game. We show the game is ordinal potential, and the existence of a pure-Nash Equilibrium (pure-NE) is therefore guaranteed. We also design a routing scheme, called Selfish Task-Driven Routing (STaR), to achieve a pure-NE. Extensive simulations show that our scheme can not only efficiently converge to an equilibrium but also outperform other source routing protocols regarding the completion time and load balancing.