Pub Date : 2025-11-24DOI: 10.1109/TNSE.2025.3636550
Fu Peng;Meng Zhang;Ming Tang
Recent studies in federated learning (FL) commonly train models on static datasets. However, real-world data often arrives as streams with shifting distributions, causing performance degradation known as concept drift. This paper analyzes FL performance under concept drift using information theory and proposes an algorithm to mitigate the performance degradation. We model concept drift as a Markov chain and introduce the Stationary Generalization Error to assess a model's capability to capture characteristics of future unseen data. Its upper bound is derived using KL divergence and mutual information. We study three drift patterns (periodic, gradual, and random) and their impact on FL performance. Inspired by this, we propose an algorithm that regularizes the empirical risk minimization approach with KL divergence and mutual information, thereby enhancing long-term performance. We also explore the performance-cost tradeoff by identifying a Pareto front. To validate our approach, we build an FL testbed using Raspberry Pi4 devices. Experimental results corroborate with theoretical findings, confirming that drift patterns significantly affect performance. Our method consistently outperforms existing approaches for these three patterns, demonstrating its effectiveness in adapting concept drift in FL.
{"title":"An Information-Theoretic Analysis for Federated Learning Under Concept Drift","authors":"Fu Peng;Meng Zhang;Ming Tang","doi":"10.1109/TNSE.2025.3636550","DOIUrl":"https://doi.org/10.1109/TNSE.2025.3636550","url":null,"abstract":"Recent studies in federated learning (FL) commonly train models on static datasets. However, real-world data often arrives as streams with shifting distributions, causing performance degradation known as concept drift. This paper analyzes FL performance under concept drift using information theory and proposes an algorithm to mitigate the performance degradation. We model concept drift as a Markov chain and introduce the <italic>Stationary Generalization Error</i> to assess a model's capability to capture characteristics of future unseen data. Its upper bound is derived using KL divergence and mutual information. We study three drift patterns (periodic, gradual, and random) and their impact on FL performance. Inspired by this, we propose an algorithm that regularizes the empirical risk minimization approach with KL divergence and mutual information, thereby enhancing long-term performance. We also explore the performance-cost tradeoff by identifying a Pareto front. To validate our approach, we build an FL testbed using Raspberry Pi4 devices. Experimental results corroborate with theoretical findings, confirming that drift patterns significantly affect performance. Our method consistently outperforms existing approaches for these three patterns, demonstrating its effectiveness in adapting concept drift in FL.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"13 ","pages":"3409-3425"},"PeriodicalIF":7.9,"publicationDate":"2025-11-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778129","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-21DOI: 10.1109/TNSE.2025.3635519
Zijun Zhan;Yaxian Dong;Daniel Mawunyo Doe;Yuqing Hu;Shuai Li;Shaohua Cao;Zhu Han
With the rapid growth in demand for AI-generated content (AIGC), edge AIGC service providers (ASPs) have become indispensable. However, designing incentive mechanisms that motivate ASPs to deliver high-quality AIGC services remains a challenge, especially in the presence of information asymmetry. In this paper, we address bonus design between a teleoperator and an edge ASP when the teleoperator cannot observe the ASP’s private settings and chosen actions (diffusion steps). We formulate this as an online learning contract design problem and decompose it into two subproblems: ASP’s settings inference and contract derivation. To tackle the NP-hard setting-inference subproblem with unknown variable sizes, we introduce a large language model (LLM)-empowered framework that iteratively refines a naive seed solver using the LLM’s domain expertise. Upon obtaining the solution from the LLM-evolved solver, we directly address the contract derivation problem using convex optimization techniques and obtain a near-optimal contract. Simulation results on our Unity-based teleoperation platform show that our method boosts the teleoperator’s utility by $5 sim 40%$ compared to benchmarks, while preserving positive incentives for the ASP.
{"title":"Learning to Incentivize: LLM-Empowered Contract for AIGC Offloading in Teleoperation","authors":"Zijun Zhan;Yaxian Dong;Daniel Mawunyo Doe;Yuqing Hu;Shuai Li;Shaohua Cao;Zhu Han","doi":"10.1109/TNSE.2025.3635519","DOIUrl":"https://doi.org/10.1109/TNSE.2025.3635519","url":null,"abstract":"With the rapid growth in demand for AI-generated content (AIGC), edge AIGC service providers (ASPs) have become indispensable. However, designing incentive mechanisms that motivate ASPs to deliver high-quality AIGC services remains a challenge, especially in the presence of information asymmetry. In this paper, we address bonus design between a teleoperator and an edge ASP when the teleoperator cannot observe the ASP’s private settings and chosen actions (diffusion steps). We formulate this as an online learning contract design problem and decompose it into two subproblems: ASP’s settings inference and contract derivation. To tackle the NP-hard setting-inference subproblem with unknown variable sizes, we introduce a large language model (LLM)-empowered framework that iteratively refines a naive seed solver using the LLM’s domain expertise. Upon obtaining the solution from the LLM-evolved solver, we directly address the contract derivation problem using convex optimization techniques and obtain a near-optimal contract. Simulation results on our Unity-based teleoperation platform show that our method boosts the teleoperator’s utility by <inline-formula><tex-math>$5 sim 40%$</tex-math></inline-formula> compared to benchmarks, while preserving positive incentives for the ASP.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"13 ","pages":"3465-3484"},"PeriodicalIF":7.9,"publicationDate":"2025-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778134","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-19DOI: 10.1109/TNSE.2025.3634750
Zhenyu Deng;Tao Zhou;Yilin Bi
Hypergraph, which allows each hyperedge to encompass an arbitrary number of nodes, is a powerful tool for modeling multi-entity interactions. Hyperedge prediction is a fundamental task that aims to predict future hyperedges or identify existing but unobserved hyperedges based on those observed. In link prediction for simple graphs, most observed links are treated as positive samples, while all unobserved links are considered as negative samples. However, this full-sampling strategy is impractical for hyperedge prediction, because to the number of unobserved hyperedges in a hypergraph significantly exceeds the number of observed ones. Therefore, one has to utilize some negative sampling methods to generate negative samples, ensuring their quantity is comparable to that of positive samples. In current hyperedge prediction, randomly selecting negative samples is a routine practice. But through experimental analysis, we discover a critical limitation of random selecting that the generated negative samples are too easily distinguishable from positive samples. This leads to premature model convergence and reduced prediction accuracy. To overcome this issue, we propose a novel method to generate negative samples, named as hard negative sampling (HNS). Unlike traditional methods that construct negative hyperedges by selecting node sets from the original hypergraph, HNS directly synthesizes negative samples in the hyperedge embedding space, thereby generating more challenging and informative negative samples. Our results demonstrate that HNS significantly enhances both accuracy and robustness of the prediction. Moreover, as a plug-and-play technique, HNS can be easily applied in the training of various hyperedge prediction models based on representation learning.
{"title":"Hard Negative Sampling in Hyperedge Prediction","authors":"Zhenyu Deng;Tao Zhou;Yilin Bi","doi":"10.1109/TNSE.2025.3634750","DOIUrl":"https://doi.org/10.1109/TNSE.2025.3634750","url":null,"abstract":"Hypergraph, which allows each hyperedge to encompass an arbitrary number of nodes, is a powerful tool for modeling multi-entity interactions. Hyperedge prediction is a fundamental task that aims to predict future hyperedges or identify existing but unobserved hyperedges based on those observed. In link prediction for simple graphs, most observed links are treated as positive samples, while all unobserved links are considered as negative samples. However, this full-sampling strategy is impractical for hyperedge prediction, because to the number of unobserved hyperedges in a hypergraph significantly exceeds the number of observed ones. Therefore, one has to utilize some negative sampling methods to generate negative samples, ensuring their quantity is comparable to that of positive samples. In current hyperedge prediction, randomly selecting negative samples is a routine practice. But through experimental analysis, we discover a critical limitation of random selecting that the generated negative samples are too easily distinguishable from positive samples. This leads to premature model convergence and reduced prediction accuracy. To overcome this issue, we propose a novel method to generate negative samples, named as hard negative sampling (HNS). Unlike traditional methods that construct negative hyperedges by selecting node sets from the original hypergraph, HNS directly synthesizes negative samples in the hyperedge embedding space, thereby generating more challenging and informative negative samples. Our results demonstrate that HNS significantly enhances both accuracy and robustness of the prediction. Moreover, as a plug-and-play technique, HNS can be easily applied in the training of various hyperedge prediction models based on representation learning.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"13 ","pages":"4833-4846"},"PeriodicalIF":7.9,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145886571","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-19DOI: 10.1109/TNSE.2025.3634548
Zihan Jiang;Qi Chen;Zihao Chen;Duncan S. Wong
Recently, coded blockchain has emerged as a key technology to address the significant storage demands resulting from the traditional blockchain’s full-replication storage mechanism. Although erasure codes can effectively reduce the storage burden on individual nodes, they introduce higher costs for data reading and repair. Moreover, most coded blockchains face challenges in adapting to dynamic networks and suffer from security vulnerabilities. In this paper, we propose DR-Store, a novel coded blockchain architecture that reduces the per-node storage cost per block from $O(n)$ to $O(1)$ by storing only a single coded block per node. DR-Store employs a reconstruction encoding scheme that minimizes the data required for decoding a single original data block, bringing it close to the size of the original block itself, thereby significantly improving read efficiency. To accommodate dynamic blockchain network environments, we introduce a reliable re-encoding process. This process allows honest nodes to either successfully complete re-encoding or safely abort it upon detecting malicious behavior from a new node, thereby securing the re-encoding procedure. Furthermore, by analyzing the Reed-Solomon re-encoding code rate as the number of nodes changes, we adaptively adjust the encoding parameters. We also propose a homomorphic re-encoding mechanism that conserves bandwidth during re-encoding, achieving faster re-encoding performance.
{"title":"DR-Store: A Dynamic Reliable Coded Blockchain Architecture","authors":"Zihan Jiang;Qi Chen;Zihao Chen;Duncan S. Wong","doi":"10.1109/TNSE.2025.3634548","DOIUrl":"https://doi.org/10.1109/TNSE.2025.3634548","url":null,"abstract":"Recently, coded blockchain has emerged as a key technology to address the significant storage demands resulting from the traditional blockchain’s full-replication storage mechanism. Although erasure codes can effectively reduce the storage burden on individual nodes, they introduce higher costs for data reading and repair. Moreover, most coded blockchains face challenges in adapting to dynamic networks and suffer from security vulnerabilities. In this paper, we propose DR-Store, a novel coded blockchain architecture that reduces the per-node storage cost per block from <inline-formula><tex-math>$O(n)$</tex-math></inline-formula> to <inline-formula><tex-math>$O(1)$</tex-math></inline-formula> by storing only a single coded block per node. DR-Store employs a reconstruction encoding scheme that minimizes the data required for decoding a single original data block, bringing it close to the size of the original block itself, thereby significantly improving read efficiency. To accommodate dynamic blockchain network environments, we introduce a reliable re-encoding process. This process allows honest nodes to either successfully complete re-encoding or safely abort it upon detecting malicious behavior from a new node, thereby securing the re-encoding procedure. Furthermore, by analyzing the Reed-Solomon re-encoding code rate as the number of nodes changes, we adaptively adjust the encoding parameters. We also propose a homomorphic re-encoding mechanism that conserves bandwidth during re-encoding, achieving faster re-encoding performance.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"13 ","pages":"3426-3443"},"PeriodicalIF":7.9,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778310","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-19DOI: 10.1109/TNSE.2025.3634598
Jing Zhang;Fei Shen;Feng Yan;Jie Li
Edge computing provides low-latency computational services for task offloading in vehicular networks. However, challenges such as dynamic transmission rates, resource limitations, and information-sharing constraints impede efficient offloading. Few studies address these issues concurrently in designing dynamic offloading strategies, often resulting in sub-optimal system utility. This paper aims to achieve efficient vehicular task offloading via an idleness-aware edge server (ES) from a game theory perspective. We propose a Gated Recurrent Unit (GRU) prediction model with an attention mechanism to guide vehicles to the nearest idle ES. The offloading decision process is modeled as a stochastic game, proving the existence of a Nash equilibrium (NE). Additionally, we model it as a multi-agent partially observable Markov decision process (POMDP) to account for limited information access among vehicles. To solve the POMDP and achieve near-optimal NE, we introduce a Multi-Agent Reinforcement Learning-based Task Offloading (MATO) algorithm, combining a Differentiable Neural Computer (DNC) and an Advantageous Actor-Critic (A2C) framework. The DNC’s external memory stores structured representations of past information, enabling deeper exploration of the strategy space. Adjusting the reward representation enhances training efficiency. Experimental results driven by real-world datasets demonstrate that MATO effectively improves the computing offloading utility while increasing the convergence speed compared to existing schemes.
{"title":"Multi-Agent Reinforcement Learning Based Idle-Aware Task Offloading in Dynamic Vehicular Networks With Partial Information","authors":"Jing Zhang;Fei Shen;Feng Yan;Jie Li","doi":"10.1109/TNSE.2025.3634598","DOIUrl":"https://doi.org/10.1109/TNSE.2025.3634598","url":null,"abstract":"Edge computing provides low-latency computational services for task offloading in vehicular networks. However, challenges such as dynamic transmission rates, resource limitations, and information-sharing constraints impede efficient offloading. Few studies address these issues concurrently in designing dynamic offloading strategies, often resulting in sub-optimal system utility. This paper aims to achieve efficient vehicular task offloading via an idleness-aware edge server (ES) from a game theory perspective. We propose a Gated Recurrent Unit (GRU) prediction model with an attention mechanism to guide vehicles to the nearest idle ES. The offloading decision process is modeled as a stochastic game, proving the existence of a Nash equilibrium (NE). Additionally, we model it as a multi-agent partially observable Markov decision process (POMDP) to account for limited information access among vehicles. To solve the POMDP and achieve near-optimal NE, we introduce a Multi-Agent Reinforcement Learning-based Task Offloading (MATO) algorithm, combining a Differentiable Neural Computer (DNC) and an Advantageous Actor-Critic (A2C) framework. The DNC’s external memory stores structured representations of past information, enabling deeper exploration of the strategy space. Adjusting the reward representation enhances training efficiency. Experimental results driven by real-world datasets demonstrate that MATO effectively improves the computing offloading utility while increasing the convergence speed compared to existing schemes.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"13 ","pages":"3499-3515"},"PeriodicalIF":7.9,"publicationDate":"2025-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778195","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Entanglement routing selects a path to establish entanglement connections between two arbitrary nodes in quantum networks, which plays an important role in quantum communication. In quantum networks, quantum decoherence and limited network performance make it challenging to distribute entangled pairs. Many entanglement routing schemes have been proposed to solve this issue but most of them are in a centralized and synchronized manner. However, they may be infeasible in large-scale quantum networks. Therefore, in this paper, we propose a distributed and asynchronous entanglement routing scheme called DFER in which quantum nodes manage requests autonomously. The major challenge is quantum nodes have little knowledge about entangled pairs, which hinders the ability to establish fidelity guaranteed entanglement connections. To address this challenge, we develop DLFR algorithm which estimates the fidelity of end-to-end entanglement connections based on link-level fidelity and calculates link-level fidelity requirement based on the characteristic of purification. Among nodes which meet link-level fidelity requirement, we design DFPS path selection algorithm to select next hop with the highest expected throughput to distribute entangled pairs. Numerous simulation results demonstrate that DFER can efficiently distribute fidelity-guaranteed entangled pairs with high throughput.
{"title":"Distributed Entanglement Routing Scheme With Fidelity Guarantee in Quantum Networks","authors":"Jiesheng Tan;Zhonghui Li;Jian Li;Bin Liu;Nenghai Yu","doi":"10.1109/TNSE.2025.3631132","DOIUrl":"https://doi.org/10.1109/TNSE.2025.3631132","url":null,"abstract":"Entanglement routing selects a path to establish entanglement connections between two arbitrary nodes in quantum networks, which plays an important role in quantum communication. In quantum networks, quantum decoherence and limited network performance make it challenging to distribute entangled pairs. Many entanglement routing schemes have been proposed to solve this issue but most of them are in a centralized and synchronized manner. However, they may be infeasible in large-scale quantum networks. Therefore, in this paper, we propose a distributed and asynchronous entanglement routing scheme called DFER in which quantum nodes manage requests autonomously. The major challenge is quantum nodes have little knowledge about entangled pairs, which hinders the ability to establish fidelity guaranteed entanglement connections. To address this challenge, we develop DLFR algorithm which estimates the fidelity of end-to-end entanglement connections based on link-level fidelity and calculates link-level fidelity requirement based on the characteristic of purification. Among nodes which meet link-level fidelity requirement, we design DFPS path selection algorithm to select next hop with the highest expected throughput to distribute entangled pairs. Numerous simulation results demonstrate that DFER can efficiently distribute fidelity-guaranteed entangled pairs with high throughput.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"13 ","pages":"3320-3334"},"PeriodicalIF":7.9,"publicationDate":"2025-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145729368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-17DOI: 10.1109/TNSE.2025.3634079
Yuanyuan Zhang;Pei Wang;Jinhu Lü;Tao Zhou
Quantifying structural dissimilarities between networks is a fundamental challenge. This paper introduces a novel measure, $D^{*}$, that quantifies network dissimilarity by analyzing their augmented networks. The augmented network is constructed by adding a virtual leader node that is bidirectionally connected to all other nodes. $D^{*}$ relies on the node distance distributions of the augmented networks, which are essentially determined by the original networks’ degree distributions and sizes. This characteristic makes it a simple and robust measure, insensitive to weighting parameters, and applicable to a wide range of networks–including regular networks and networks of any size, even those containing isolated nodes. $D^{*}$ actually utilizes the second-order truncated shortest-path distance matrix and demonstrates superior performance compared to higher-order truncations. Numerical simulations show that $D^{*}$ accurately quantifies structural differences between networks while overcoming the saturation growth effect induced by increasing edge connection probabilities in random networks. The versatility of $D^{*}$ is further demonstrated by applying it to four distinct scenarios. Specifically, $D^{*}$ is effective in determining optimal correlation cutoff thresholds when constructing bio-molecular co-expression networks, in identifying disease modules in gene-disease and phenotype-disease networks, in performing clustering and layer aggregation for multilayer networks, and in distinguishing networks of different categories. Our findings enhance understanding of the construction and comparison of complex network structures.
{"title":"Quantifying Network Dissimilarity via Augmented Networks","authors":"Yuanyuan Zhang;Pei Wang;Jinhu Lü;Tao Zhou","doi":"10.1109/TNSE.2025.3634079","DOIUrl":"https://doi.org/10.1109/TNSE.2025.3634079","url":null,"abstract":"Quantifying structural dissimilarities between networks is a fundamental challenge. This paper introduces a novel measure, <inline-formula><tex-math>$D^{*}$</tex-math></inline-formula>, that quantifies network dissimilarity by analyzing their augmented networks. The augmented network is constructed by adding a virtual leader node that is bidirectionally connected to all other nodes. <inline-formula><tex-math>$D^{*}$</tex-math></inline-formula> relies on the node distance distributions of the augmented networks, which are essentially determined by the original networks’ degree distributions and sizes. This characteristic makes it a simple and robust measure, insensitive to weighting parameters, and applicable to a wide range of networks–including regular networks and networks of any size, even those containing isolated nodes. <inline-formula><tex-math>$D^{*}$</tex-math></inline-formula> actually utilizes the second-order truncated shortest-path distance matrix and demonstrates superior performance compared to higher-order truncations. Numerical simulations show that <inline-formula><tex-math>$D^{*}$</tex-math></inline-formula> accurately quantifies structural differences between networks while overcoming the saturation growth effect induced by increasing edge connection probabilities in random networks. The versatility of <inline-formula><tex-math>$D^{*}$</tex-math></inline-formula> is further demonstrated by applying it to four distinct scenarios. Specifically, <inline-formula><tex-math>$D^{*}$</tex-math></inline-formula> is effective in determining optimal correlation cutoff thresholds when constructing bio-molecular co-expression networks, in identifying disease modules in gene-disease and phenotype-disease networks, in performing clustering and layer aggregation for multilayer networks, and in distinguishing networks of different categories. Our findings enhance understanding of the construction and comparison of complex network structures.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"13 ","pages":"3566-3579"},"PeriodicalIF":7.9,"publicationDate":"2025-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778262","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Federated learning (FL) facilitates decentralized machine learning by enabling participating entities to cooperatively train shared models while retaining local data ownership. To address privacy concerns inherent in distributed training, privacy-preserving mechanisms have been incorporated into FL frameworks to enhance confidentiality and protect sensitive data. However, the implementation of these privacy-enhancing techniques introduces vulnerabilities to poisoning attacks, wherein malicious actors manipulate training processes to degrade model integrity or performance. Current defense strategies rely on statistical methods to mitigate such attacks, but they lack sufficient robustness, demonstrating limited effectiveness against diverse attack types or scenarios where malicious clients exceed a small minority. To overcome these limitations, we propose RobustPPFL, a privacy-preserving federated learning framework designed to withstand multiple poisoning attack types with high resilience, even when malicious participants dominate the client population. Our approach integrates three core innovations. First, we add performance verification of client-encrypted models during the training process to detect malicious clients in-time. Second, a secure model inference protocol is proposed to enable privacy-preserving training. Last but not the least, we design a grouped verification mechanism enhanced by hierarchical aggregation rules to optimize efficiency and minimize interference in malicious client detection. We evaluate RobustPPFL through extensive experiments across diverse datasets and attack scenarios. The experimental results show that our proposed framework achieves privacy preservation and it achieves high robustness against poisoning attacks.
{"title":"RobustPPFL: A Secure and Robust Privacy-Preserving Federated Learning Framework Against Poisoning Attacks","authors":"Kaiping Xue;Jiachen Li;Rui Xue;Yingjie Xue;Jingcheng Zhao","doi":"10.1109/TNSE.2025.3632902","DOIUrl":"https://doi.org/10.1109/TNSE.2025.3632902","url":null,"abstract":"Federated learning (FL) facilitates decentralized machine learning by enabling participating entities to cooperatively train shared models while retaining local data ownership. To address privacy concerns inherent in distributed training, privacy-preserving mechanisms have been incorporated into FL frameworks to enhance confidentiality and protect sensitive data. However, the implementation of these privacy-enhancing techniques introduces vulnerabilities to poisoning attacks, wherein malicious actors manipulate training processes to degrade model integrity or performance. Current defense strategies rely on statistical methods to mitigate such attacks, but they lack sufficient robustness, demonstrating limited effectiveness against diverse attack types or scenarios where malicious clients exceed a small minority. To overcome these limitations, we propose RobustPPFL, a privacy-preserving federated learning framework designed to withstand multiple poisoning attack types with high resilience, even when malicious participants dominate the client population. Our approach integrates three core innovations. First, we add performance verification of client-encrypted models during the training process to detect malicious clients in-time. Second, a secure model inference protocol is proposed to enable privacy-preserving training. Last but not the least, we design a grouped verification mechanism enhanced by hierarchical aggregation rules to optimize efficiency and minimize interference in malicious client detection. We evaluate RobustPPFL through extensive experiments across diverse datasets and attack scenarios. The experimental results show that our proposed framework achieves privacy preservation and it achieves high robustness against poisoning attacks.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"13 ","pages":"3351-3368"},"PeriodicalIF":7.9,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778281","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-14DOI: 10.1109/TNSE.2025.3632547
Zhongqiang Zhang;Shuhang Zhang;Haoyong Li;Guangming Shi;Jiayin Xue;Bin Li
Remote sensing images play a crucial and indispensable role in many fields such as environmental monitoring and geological disaster detection. With the advancement of satellite remote sensing acquisition technology, the remote sensing data shows explosive growth. However, the current communication bandwidth of the satellite-to-ground channel is difficult to meet the requirements for massive remote sensing data transmission. To this end, this paper proposes an elastic coding and decoding method for remote sensing image semantic transmission. The proposed transmission method includes a global-local feature extraction module, a key semantic feature selection module, a joint source channel coding module, a decoding module, and an analysis module. The global-local feature extraction module can effectively extract global context features and local detailed features via multi-directional mamba block and residual block, respectively. The key semantic feature selection module can elastically select key features according to channel state signal-to-noise ratios (SNRs). The joint source channel coding and decoding modules can further improve the transmission robustness via adding different types of channel conditions. The proposed method only needs to transmit key information in remote sensing images while discarding the redundant information, which significantly improves the transmission efficiency of remote sensing images. The extensive experimental results on the NWPU-RESISC45, UCMerced-LandUse, AID, and RSSCN7 datasets demonstrate that our method obtains higher transmission accuracies and transmission efficiency than state-of-the-art methods.
{"title":"An Elastic Coding and Decoding Method for Satellite Remote Sensing Image Semantic Transmission","authors":"Zhongqiang Zhang;Shuhang Zhang;Haoyong Li;Guangming Shi;Jiayin Xue;Bin Li","doi":"10.1109/TNSE.2025.3632547","DOIUrl":"https://doi.org/10.1109/TNSE.2025.3632547","url":null,"abstract":"Remote sensing images play a crucial and indispensable role in many fields such as environmental monitoring and geological disaster detection. With the advancement of satellite remote sensing acquisition technology, the remote sensing data shows explosive growth. However, the current communication bandwidth of the satellite-to-ground channel is difficult to meet the requirements for massive remote sensing data transmission. To this end, this paper proposes an elastic coding and decoding method for remote sensing image semantic transmission. The proposed transmission method includes a global-local feature extraction module, a key semantic feature selection module, a joint source channel coding module, a decoding module, and an analysis module. The global-local feature extraction module can effectively extract global context features and local detailed features via multi-directional mamba block and residual block, respectively. The key semantic feature selection module can elastically select key features according to channel state signal-to-noise ratios (SNRs). The joint source channel coding and decoding modules can further improve the transmission robustness via adding different types of channel conditions. The proposed method only needs to transmit key information in remote sensing images while discarding the redundant information, which significantly improves the transmission efficiency of remote sensing images. The extensive experimental results on the NWPU-RESISC45, UCMerced-LandUse, AID, and RSSCN7 datasets demonstrate that our method obtains higher transmission accuracies and transmission efficiency than state-of-the-art methods.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"13 ","pages":"3485-3498"},"PeriodicalIF":7.9,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778328","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-14DOI: 10.1109/TNSE.2025.3632865
Bingqing Ren;Peng Yang;Miao Du;Dongmei Yang
Multi-access edge computing (MEC) is an important technology to accelerate the response speed of computation-intensive and delay-sensitive tasks, which promotes the development of Artificial Intelligence (AI) and Internet of Things (IoT). Federated learning (FL) over mobile devices, coupled with MEC to build an intelligent network, can avoid the risk of privacy leakage by keeping the sensitive information of each agent locally. However, there exist some problems in implementing FL over mobile devices, such as the shortage of edge bandwidth and computing resources. In addition, network dynamics and the heterogeneity of mobile devices need to be considered. To address these issues, we propose a heterogeneous Stackelberg game approach based on deep reinforcement learning (DRL) to achieve the desired trade-off between computing and communication in the FL system, called Energy Efficient Heterogeneous Federated Learning (EEHFL). Specifically, EEHFL designs a new two-stage Stackelberg game approach based on the heterogeneous FL architecture with convergence guarantee targeting efficient energy, which is modeled separately. Furthermore, DRL algorithms are introduced to solve the problem, realizing the control of heterogeneous parameters in a dynamic environment. Experimental results illustrate that compared with state-of-the-art baselines, our model achieves remarkable improvement, which demonstrates the superiority of our model on saving cost and energy consumption.
{"title":"Energy Efficient Heterogeneous Federated Learning Over Mobile Devices: A Deep Reinforcement Learning Based Stackelberg Game Approach","authors":"Bingqing Ren;Peng Yang;Miao Du;Dongmei Yang","doi":"10.1109/TNSE.2025.3632865","DOIUrl":"https://doi.org/10.1109/TNSE.2025.3632865","url":null,"abstract":"Multi-access edge computing (MEC) is an important technology to accelerate the response speed of computation-intensive and delay-sensitive tasks, which promotes the development of Artificial Intelligence (AI) and Internet of Things (IoT). Federated learning (FL) over mobile devices, coupled with MEC to build an intelligent network, can avoid the risk of privacy leakage by keeping the sensitive information of each agent locally. However, there exist some problems in implementing FL over mobile devices, such as the shortage of edge bandwidth and computing resources. In addition, network dynamics and the heterogeneity of mobile devices need to be considered. To address these issues, we propose a heterogeneous Stackelberg game approach based on deep reinforcement learning (DRL) to achieve the desired trade-off between computing and communication in the FL system, called Energy Efficient Heterogeneous Federated Learning (EEHFL). Specifically, EEHFL designs a new two-stage Stackelberg game approach based on the heterogeneous FL architecture with convergence guarantee targeting efficient energy, which is modeled separately. Furthermore, DRL algorithms are introduced to solve the problem, realizing the control of heterogeneous parameters in a dynamic environment. Experimental results illustrate that compared with state-of-the-art baselines, our model achieves remarkable improvement, which demonstrates the superiority of our model on saving cost and energy consumption.","PeriodicalId":54229,"journal":{"name":"IEEE Transactions on Network Science and Engineering","volume":"13 ","pages":"3534-3550"},"PeriodicalIF":7.9,"publicationDate":"2025-11-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145778196","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：