Computer Networks最新文献

{"title":"Multi-policy reinforcement learning for network resource allocation with periodic behaviors","authors":"Zheyu Chen ,&nbsp;Kin K. Leung ,&nbsp;Shiqiang Wang ,&nbsp;Leandros Tassiulas ,&nbsp;Kevin Chan ,&nbsp;Patrick J. Baker","doi":"10.1016/j.comnet.2025.111645","DOIUrl":"10.1016/j.comnet.2025.111645","url":null,"abstract":"<div><div>Markov Decision Processes (MDPs) serve as the mathematical foundation of Reinforcement learning (RL), where a Markov process with defined states is used to model the system and the actions to be taken affect the state transitions and the corresponding rewards. The RL and deep RL (DRL) can produce the high-performing action policy to maximize the long-term reward. Although RL/DRL have been widely applied to communication and computer systems, a key limitation is that the system under consideration often does not satisfy the required mathematical properties, thus making the MDP inexact and the derived policy flawed. Therefore, we consider the periodic Markov Decision Process (pMDP), where the evolution of the underlying process and model parameters for the pMDP demonstrate some forms of periodic characteristics (e.g., periodic job arrivals and available resources) which violate the Markov property. To obtain the optimal policies for the pMDP, a policy gradient method with a multi-policy solution framework is proposed, and a deep-learning method is developed to improve the effectiveness and stability of the proposed solution. Furthermore, a layer-sharing strategy is proposed to reduce the storage complexity by reducing the number of parameters in the neural networks. The deep-learning method is applied to achieve the near-optimal allocation of resources to arriving computational tasks in a network setting corresponding to the software-defined network (SDN). Evaluation results reveal that the proposed technique is valid and capable of outperforming a baseline method that employs a single policy by 31% on average.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"272 ","pages":"Article 111645"},"PeriodicalIF":4.6,"publicationDate":"2025-08-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144920039","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}

{"title":"Alternating optimization for energy consumption-oriented task offloading in SAGIN","authors":"Pengfei Yang ,&nbsp;Tianyang Zheng ,&nbsp;Shuyu Zhang ,&nbsp;Weidi Su ,&nbsp;Bijie Yi ,&nbsp;Wenkai Lv ,&nbsp;Quan Wang","doi":"10.1016/j.comnet.2025.111646","DOIUrl":"10.1016/j.comnet.2025.111646","url":null,"abstract":"<div><div>As a fundamental framework for future mobile communication systems, the Space-Air-Ground Integrated Network (SAGIN) leverages Multi-Access Edge Computing (MEC) technology to provide communication and computation resources to users by connecting devices at different network levels. However, current task offloading schemes often suffer from suboptimal energy consumption performance and fail to balance various constraints, such as communication resources, computation resources, maximum allowable task delay, and the maximum coverage time of Low Earth orbit (LEO) satellites. To address these issues, this paper focuses on the energy consumption-driven task offloading problem within SAGIN. Our approach differs from existing models by explicitly considering the aforementioned constraints in the task allocation process. To tackle the complexity of the original problem more efficiently and to provide a structured way to address both the communication and computation processes, the problem is formulated into two subproblems: the bandwidth allocation problem and the MEC task offloading decision problem. To solve these subproblems, we propose the Joint Bandwidth Allocation and MEC Task Offloading decision-alternating Optimization (JBAMTO-AO) algorithm. The CVX toolkit and the alternating direction method of multipliers (ADMM) distributed algorithm are employed to address the decomposed subproblems effectively. Extensive experimental evaluations show that the proposed JBAMTO-AO algorithm outperforms existing offloading methods in terms of both the total energy consumption of system tasks and the satisfaction degree of task maximum tolerable delay.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"271 ","pages":"Article 111646"},"PeriodicalIF":4.6,"publicationDate":"2025-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144896311","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}

{"title":"SCEP-TI: A side-channel attack on encrypted proxy video streams for video title identification","authors":"Yi Zhang ,&nbsp;Zhenyu Xu ,&nbsp;Xurui Ren ,&nbsp;Hua Wu ,&nbsp;Guang Cheng","doi":"10.1016/j.comnet.2025.111630","DOIUrl":"10.1016/j.comnet.2025.111630","url":null,"abstract":"<div><div>With the widespread use of the Internet and the continuous development of streaming media technology, video streaming has increasingly become the main body of the entire Internet traffic. Although the video is encrypted during transmission, it may still be vulnerable to side-channel attacks. However, after the video is transmitted through proxy encapsulation, the accuracy of side-channel attack is greatly reduced. In the paper, we propose a side-channel attack on encrypted proxy video streams for title identification(SCEP-TI), which is a new attack method to identify video titles in encrypted proxy traffic based on side-channel features. We extract stable video segment features from encrypted proxy traffic based on DASH and HLS protocols, and SCEP-TI utilizes a convolutional neural network (CNN) model to accurately identify video titles. This method overcomes the traffic confusion caused by encrypted proxy encapsulation, the interference of a large amount of background traffic, and the limitation of unidirectional traffic in asymmetric routing scenarios. The experimental results show that SCEP-TI has higher accuracy in both closed-world and open-world scenarios, which is superior to the existing methods. Furthermore, to protect the privacy of the user, we propose two defense mechanisms. Our code is available at <span><span>https://github.com/Zzzyyzz/SCEP-TI</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"271 ","pages":"Article 111630"},"PeriodicalIF":4.6,"publicationDate":"2025-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144896310","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}

{"title":"EMTD: Efficient encrypted malware traffic detection based on adaptive meta-path guided graph propagation","authors":"Fanyi Zeng,&nbsp;Dapeng Man,&nbsp;Yuhao Zhao,&nbsp;Yuchen Liu,&nbsp;Huanran Wang,&nbsp;Wu Yang","doi":"10.1016/j.comnet.2025.111636","DOIUrl":"10.1016/j.comnet.2025.111636","url":null,"abstract":"<div><div>Given the growing challenges posed by encrypted Android malware, developing effective detection methods is crucial to understanding the evolution of malware families and designing preventive security measures. Existing detection methods for encrypted malware traffic primarily focus on extracting features at the single-flow level and multi-flow context level, failing to capture the evolutionary associations within known malware families and their previously unseen variants, which compromises detection effectiveness. Graph-based modeling methods have advantages in expressing traffic association features, but scalability issues present new challenges for detection timeliness. To address these limitations, we propose a novel two-stage detection framework named <strong>Encrypted Malware Traffic Detection (EMTD)</strong>. In the first phase, our method <strong>MFGDect</strong> models encrypted traffic as a heterogeneous information network and applies a multilayer heterogeneous attention mechanism to learn semantic associations among traffic flows. This enables adaptive family-aware representation and improves detection performance under encryption. Furthermore, we design <strong>MFGDect++</strong>, an extension of our base model that introduces adaptive meta-path guided graph propagation, enabling efficient incremental detection of new traffic samples without re-graphing or model retraining. This mechanism significantly reduces the average detection time to 135 ms per sample, demonstrating strong scalability. Experiments on public datasets demonstrate that EMTD outperforms existing baselines, achieving an average improvement of 9.62% in malicious sample recall and a 2.32% increase in F1 score, while maintaining low resource overheads and strong adaptability to large-scale graph data.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"271 ","pages":"Article 111636"},"PeriodicalIF":4.6,"publicationDate":"2025-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144902884","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}

{"title":"Real-time adaptive resource management for high-resolution computer vision over private 5G networks","authors":"Rui Silva,&nbsp;Filipe Antão,&nbsp;David Santos,&nbsp;André Perdigão,&nbsp;Tiago Barros,&nbsp;Fatma Marzouk,&nbsp;Alisson Chaves,&nbsp;Daniel Corujo,&nbsp;Rui L. Aguiar","doi":"10.1016/j.comnet.2025.111644","DOIUrl":"10.1016/j.comnet.2025.111644","url":null,"abstract":"<div><div>The advancements in wireless networking technologies empower new opportunities and applications. However, these advanced networks are increasingly consuming more energy to provide higher performance. In line with the United Nations sustainable development goals and the need to reduce networking energy consumption, this paper presents an efficient dynamic slicing and antenna control architecture for 5G networks. The architecture was then applied to a smart port scenario, where a pre-gate is used to control the entrance of trucks into the port with the aid of a camera, with the video stream being analyzed in a datacenter to detect the presence of a truck and access its details. Experimental results showed that the architecture was able to potentially reduce the energy expenditure of the considered scenario in 2.26 MWh in a year, considering a real statistical number of trucks entering the Portuguese Sines Port in 2017. Moreover, it reduced the energy consumption of the User Equipment, computing infrastructure, and wireless network in 9%, 23% and 14% respectively.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"271 ","pages":"Article 111644"},"PeriodicalIF":4.6,"publicationDate":"2025-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144896312","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}

{"title":"Reliability-aware hybrid SFC backup and deployment in edge computing","authors":"Yue Zeng ,&nbsp;Pan Li ,&nbsp;Shanshan Lin ,&nbsp;Bin Tang ,&nbsp;Xiaoliang Wang ,&nbsp;Zhihao Qu ,&nbsp;Baoliu Ye ,&nbsp;Song Guo ,&nbsp;Junlong Zhou","doi":"10.1016/j.comnet.2025.111641","DOIUrl":"10.1016/j.comnet.2025.111641","url":null,"abstract":"<div><div>As key enabling technologies for 5G, network function virtualization (NFV) abstracts services into software-based network function chains called service function chains (SFCs), greatly simplifying service management, while edge computing pushes compute resources to the edge close to IoT users, enabling low-latency services. For mission-critical applications, backup is an effective way to enhance the reliability of deployed SFCs. However, existing backup and deployment schemes mainly focus on off-site backups, neglecting on-site backups and resulting in suboptimal solutions. This paper investigates the problem of reliable SFC hybrid backup and deployment, aiming to minimize resource costs while accounting for limited edge resources and the heterogeneity of software reliability, hardware reliability, and resource charging. To tackle this problem, we first establish the mathematical association between backup and deployment decisions and these factors, formalize it as an integer nonlinear programming problem, and analyze its complexity. Then, we devise a bi-criteria approximation algorithm with rigorous theoretical guarantees, which relaxes the formalized problem to a convex optimization and rounds the fractional solution obtained by solving this convex optimization based on our insight, which approaches the optimal solution with bounded resource capacity violation, and is suitable for scenarios where moderate resource over-allocation is allowed. For cases prohibiting over-allocation, we propose a priority-guided algorithm with rigorous theoretical guarantees based on our insights, which prioritizes deploying backups for VNFs with the lowest reliability on edge sites that bring higher reliability improvements and lower charges. Extensive evaluation results show that our algorithm can save costs by up to 60.3% compared with state-of-the-art solutions.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"271 ","pages":"Article 111641"},"PeriodicalIF":4.6,"publicationDate":"2025-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144878782","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}

{"title":"Optimized edge-cloud task offloading for WBANs: A hierarchical deep-reinforcement-learning approach","authors":"Heba M. Khater ,&nbsp;Farag Sallabi ,&nbsp;Abdulmalik Alwarafy ,&nbsp;Ezedin Barka ,&nbsp;Mohamed Adel Serhani ,&nbsp;Khaled Shuaib ,&nbsp;Mohamad Khayat","doi":"10.1016/j.comnet.2025.111640","DOIUrl":"10.1016/j.comnet.2025.111640","url":null,"abstract":"<div><div>The emergence of wearable medical devices and wireless body area networks (WBANs) has enabled continuous, real-time patient monitoring. These systems generate large volumes of health data, requiring low-latency and reliable processing for timely interventions. However, local processing is often inefficient due to the energy and computational limitations of mobile devices. Offloading tasks to edge computing and cloud resources offers a promising alternative. Nonetheless, optimizing offloading decisions in dynamic healthcare scenarios remains challenging due to heterogeneous task requirements and varying computational resources. This paper presents a hierarchical actor-critic task offloading approach (HACTO), a deep-reinforcement-learning framework designed to enhance the efficiency and adaptability of task offloading in healthcare scenarios. By introducing a hierarchical decision structure, HACTO reduces complexity and improves learning performance. The problem is modeled as a Markov decision process and solved using the deep deterministic policy gradient algorithm. HACTO jointly optimizes task offloading with respect to three objectives: meeting task deadlines, minimizing the energy consumption of mobile devices, and reducing resource usage costs. Our experimental results show that HACTO outperforms traditional and deep-reinforcement-learning-based offloading strategies, making it a promising solution for intelligent task offloading in resource-constrained WBAN environments.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"271 ","pages":"Article 111640"},"PeriodicalIF":4.6,"publicationDate":"2025-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144886486","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}

{"title":"Federated resource prediction in UAV networks for efficient composition of drone delivery services","authors":"Haithem Mezni ,&nbsp;Mokhtar Sellami ,&nbsp;Hela Elmannai ,&nbsp;Reem Alkanhel","doi":"10.1016/j.comnet.2025.111642","DOIUrl":"10.1016/j.comnet.2025.111642","url":null,"abstract":"<div><div>As drones continue to see widespread adoption across commercial, private, healthcare, and education sectors, their commercial use is experiencing rapid growth. To enhance drone-based package delivery efficiency and improve customer experience, the vast amount of flight and recharging data collected from drones and stations offers valuable opportunities for predicting both resource availability within the sky network and drones’ capacity to complete delivery missions. However, the variations in regional regulations and privacy restrictions enforced by drone service providers lead to data heterogeneity, making centralized processing of flight and charging history impractical. Running machine learning models locally at the service provider level (drones and stations) addresses privacy concerns, yet processing the large volume and diversity of raw data remains a significant challenge. To deal with these issues, a collaborative learning approach based on historical delivery data presents an elegant solution. Aiming to offer predictive scheduling of drone delivery missions, while taking into consideration the complexity, heterogeneity, and dynamic nature of their flight environment, we propose a predictive and federated approach for the resilient selection and composition of drone delivery services, leveraging the strengths of federated learning (FL) to handle data privacy and heterogeneity. Our method utilizes a federated Recurrent Neural Network (FL-RNN) model that combines predictive capabilities with federated behavior, enabling collaborative forecasting and efficient mission scheduling in low-congestion regions based on the most reliable drone services. Additionally, an enhanced A* search algorithm is defined to identify the optimal delivery path by factoring in station overload probabilities. Besides computational efficiency, experimental results demonstrate the effectiveness of our approach, achieving a 16.66% improvement in prediction accuracy and an 8.89% reduction in delivery costs compared to non-federated and non-predictive solutions.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"271 ","pages":"Article 111642"},"PeriodicalIF":4.6,"publicationDate":"2025-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144886490","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}

{"title":"A deep learning-based reverse auction mechanism for semantic communication in IoV crowdsensing services","authors":"Peng Chen ,&nbsp;Youtong Li ,&nbsp;Hao Wu ,&nbsp;Jixian Zhang","doi":"10.1016/j.comnet.2025.111643","DOIUrl":"10.1016/j.comnet.2025.111643","url":null,"abstract":"<div><div>Internet of Vehicles (IoV) crowdsensing is an efficient approach to vehicle data collection in which vehicle service providers (VSPs) recruit users to participate in IoV crowdsensing tasks to obtain large amounts of vehicle data at low costs. However, the massive amount of vehicular data imposes significant challenges to the limited storage and communication resources, thereby hindering the efficient acquisition of the required information. To address these challenges, this paper proposes multiple effective strategies. To address the challenge of large data volumes, we employ semantic communication techniques to effectively compress the collected data for efficient storage and transmission. Furthermore, we define a semantic information value function to quantify the value of vehicular semantic information, and to address the problem of slow data transmission, we propose shunting offloading data to edge servers to improve the transmission efficiency. Building on this foundation, we design a deep learning-based reverse auction mechanism, SVRANet, to effectively allocate crowdsensing tasks and communication resources. SVRANet leverages self-attention mechanisms to uncover latent interactions within the information, thereby enhancing the model’s ability to allocate resources more efficiently. The experimental results demonstrate that SVRANet performs excellently, achieving high utility and social welfare while guaranteeing incentive compatibility, individual rationality, and budget feasibility.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"271 ","pages":"Article 111643"},"PeriodicalIF":4.6,"publicationDate":"2025-08-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144886489","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}

{"title":"Temporal graph attention and contrastive learning model for link prediction in dynamic networks","authors":"Feng Zhang ,&nbsp;Chenhao Luo ,&nbsp;Winston K.G. Seah ,&nbsp;Gang Xu ,&nbsp;Kailiang Zhao","doi":"10.1016/j.comnet.2025.111596","DOIUrl":"10.1016/j.comnet.2025.111596","url":null,"abstract":"<div><div>Existing discrete time-slicing methods suffer from three critical limitations: coarse temporal processing granularity, exclusively model connection establishment events while neglecting the impact of disconnection events, and vulnerability to sample imbalance and transient connection noise. These limitations severely constrain adaptability in dynamic networks characterized by connection instability and interaction volatility, where both connection establishment and disconnection events govern topological evolution. To address these challenges, this paper proposes a novel dynamic link prediction framework integrating Contrastive Learning with enhanced Temporal Graph Attention Network (CLTGAT). Our model employs continuous timestamp encoding to explicitly incorporate connection moments while fusing disconnection moments via weighted mechanisms. This dual-event modelling enables joint influence of connection/disconnection timestamps on link states. Crucially, we design a connection-duration-based Top-K contrastive sampling strategy to simultaneously mitigate transient connection noise and sample imbalance, selecting stable neighbours as positive samples while controlling distribution to alleviate prediction bias. Evaluations on three highly volatile real-world dynamic network datasets, viz., Infocom05, Hyccups and Infocom06, demonstrate CLTGAT’s superiority. Compared with seven other methods, our approach achieves higher link prediction accuracy with reduced training time, exhibiting enhanced adaptability to rapidly evolving network scenarios.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"271 ","pages":"Article 111596"},"PeriodicalIF":4.6,"publicationDate":"2025-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144886485","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}