Computer Networks最新文献

{"title":"GSDRL: Resource orchestration in integrated satellite–terrestrial network for a hybrid edge–cloud","authors":"Hongxia Zhang,&nbsp;Ning Gao,&nbsp;Xiangxu Zhao,&nbsp;Yuanbo Zhao,&nbsp;Peiying Zhang","doi":"10.1016/j.comnet.2025.111403","DOIUrl":"10.1016/j.comnet.2025.111403","url":null,"abstract":"<div><div>Integrated satellite terrestrial network (ISTN) has the potential to enable extensive global communication coverage, and resource orchestration is a hot issue in ISTN research. Network function virtualisation (NFV) technology enables the construction of service function chains (SFCs), which support flexible and efficient heterogeneous resource management. However, satellite–terrestrial edge nodes face resource constraints, making it challenging to meet growing service demands. Therefore, this paper proposes an edge–cloud collaboration-based ISTN resource orchestration method to achieve efficient service provisioning. Specifically, first, an edge–cloud hybrid oriented ISTN management architecture is established. Next, the SFC orchestration problem is transformed into a joint optimization problem of service acceptance and service cost, while satellite mobility is taken into account to ensure reliable service. Then, we propose a deep reinforcement learning-based SFC orchestration algorithm, GSDRL, which utilizes a Graph Attention Network-based feature extraction module to identify the intrinsic dependencies in the network topology. Additionally, we design an action mapping network with a coding scheme for sequence-to-sequence generation of VNF deployment decisions. Finally, extensive simulation experiments show that the proposed approach improves the service acceptance rate by 18.6% and reduces the average service cost by 15.6% compared to the benchmark solution.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"269 ","pages":"Article 111403"},"PeriodicalIF":4.4,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144279766","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":"Intrusion detection in Software-Defined Networking using hybrid Bayesian model averaging for reliable uncertainty quantification","authors":"Yasmine Medjadba ,&nbsp;Hamza Drid ,&nbsp;Mohamed Rahouti","doi":"10.1016/j.comnet.2025.111436","DOIUrl":"10.1016/j.comnet.2025.111436","url":null,"abstract":"<div><div>Managing uncertainty in anomaly detection within Software-Defined Networking (SDN) is key to maintaining resilient security and enabling effective decision-making. The SDN controller, which serves as the network’s central hub, is highly susceptible to attacks, posing considerable security challenges. This inherent vulnerability undermines the precise detection of potential threats, often resulting in a higher rate of false positives. Consequently, this compromises the trustworthiness of Intrusion Detection Systems (IDS). This study introduces a novel method to improve predictive accuracy and uncertainty quantification using Bayesian ensembles. Our method leverages Convolutional Neural Networks and Long Short-Term Memory (CNN-LSTM) with Monte Carlo (MC) Dropout to assess predictive uncertainty. Using a Bayesian approach, uncertainty is quantified by averaging predictions and calculating the standard deviation across multiple stochastic forward passes within the ensemble. Experimental results on the InSDN and CIC-DDoS2019 datasets demonstrate a significant improvement in performance and more reliable uncertainty estimates compared to state-of-the-art methods.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"269 ","pages":"Article 111436"},"PeriodicalIF":4.4,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144240467","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":"Libra: High-precision congestion control for datacenter networks with in-network allreduce","authors":"Guannan Zhang,&nbsp;Dezun Dong,&nbsp;Tao Li,&nbsp;Siyuan Yang","doi":"10.1016/j.comnet.2025.111407","DOIUrl":"10.1016/j.comnet.2025.111407","url":null,"abstract":"<div><div>Allreduce is an essential collective communication operation in distributed and parallel computing. The most advanced In-Network Allreduce (INA) technique accelerates this process by offloading the aggregation function onto the network infrastructure. However, in data center networks, allreduce traffic often coexists with background traffic, creating serious challenges for congestion control. Current INA Congestion Control (CC) systems rely on path-level signals to detect network congestion, which struggles to effectively manage mixed traffic patterns and leads to performance bottlenecks.</div><div>We propose Libra, a congestion management system for datacenter networks with INA, which aims to provide low latency and high-bandwidth services for the overall traffic. Libra utilizes the switch’s in-network telemetry feature to acquire precise link data and meticulously adjusts the congestion window via a tailored synchronization protocol. This allows Libra to efficiently adapt to intricate traffic patterns without losing network utilization. Extensive simulations demonstrate that Libra outperforms advanced approaches by 1.23x on the execution efficiency of allreduce, as well as creates significant performance gains for background traffic.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"269 ","pages":"Article 111407"},"PeriodicalIF":4.4,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144288749","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":"Self-similar traffic prediction algorithm for satellite network based on dual decomposition and neural network","authors":"Yuxia Bie ,&nbsp;Xin Li ,&nbsp;Ye Tian ,&nbsp;Yupeng Wang","doi":"10.1016/j.comnet.2025.111432","DOIUrl":"10.1016/j.comnet.2025.111432","url":null,"abstract":"<div><div>To treat the self-similar and long-term correlation characteristics of satellite network service traffic for data transmission, a traffic prediction model is proposed combining dual decomposition and Convolutional Neural Network-Bi-directional Long Short-term Memory (CNN-BiLSTM) model. First, first-mode decomposition is applied to the original self-similar traffic sequence of the satellite network using the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN). Second, the complexity of the data is evaluated by calculating the Sample Entropy (SE) of each component obtained by decomposition. The K-means clustering algorithm is used to classify the components into three groups, namely, high- frequency component, medium-frequency component and low-frequency component. The high-frequency component is subjected to secondary Variational Mode decomposition (VMD). Finally, the components obtained by the two-stage decomposition are input into the CNN-BiLSTM to yield the predicted traffic. Simulation results show that the traffic prediction model that integrates dual decomposition with CNN-BiLSTM enhances both the accuracy and efficiency of self-similar traffic predictions, thereby boosting the stability and reliability of satellite network.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"269 ","pages":"Article 111432"},"PeriodicalIF":4.4,"publicationDate":"2025-06-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144254407","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":"Personalized federated learning through self-knowledge distillation in vehicular edge computing","authors":"Gengjian Liao,&nbsp;Yulin Yang,&nbsp;Zhenni Feng","doi":"10.1016/j.comnet.2025.111406","DOIUrl":"10.1016/j.comnet.2025.111406","url":null,"abstract":"<div><div>Cooperative Vehicle-Infrastructure System (CVIS) technology enabling the exchange of information between road infrastructure and vehicles, has emerged as a notable research direction from the public. Meanwhile, the demand for personalized services tailored to individual vehicles is increasingly prominent. Under challenges such as data heterogeneity and unstable communication among vehicle clients, achieving collaborative computing that meets the personalized needs of each vehicle client has become a significant challenge. In this paper, we present pFedVS, a personalized federated learning(PFL) approach under CVIS. On one hand, we design a novel evaluation metric and algorithm based on client contribution to assist the server in client selection, aiming to mitigate the impact of communication stability and fairness on personalized performance across vehicle clients. On the other hand, considering that the performance of personalized models on vehicle clients may degrade when not selected for a long period, we adopt self-distillation for personalized model training, which not only satisfies the personalized needs of each vehicle client but also reduces computational overhead.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"269 ","pages":"Article 111406"},"PeriodicalIF":4.4,"publicationDate":"2025-06-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144254408","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":"Decentralized reinforcement learning-based resource management for dense-interference IoV","authors":"Chengbo Yu ,&nbsp;Yidan Sun ,&nbsp;Xingxing Li ,&nbsp;Fan Yang ,&nbsp;Jie Huang ,&nbsp;Yingying Peng","doi":"10.1016/j.comnet.2025.111402","DOIUrl":"10.1016/j.comnet.2025.111402","url":null,"abstract":"<div><div>In the Internet of Vehicles (IoV) with dense interference, overlapping coverage between vehicle user equipments’ communication radius (VUEs) leads to significant interference, causing resource reuse conflicts between transmission links. To address this, we construct an interference hypergraph model to analyze the interference relationships among multiple users through hyperedges. The degree of interference for the overall network is calculated using this model. The optimization problem for resource allocation is then formulated to minimize the data rate so that multiple quality of service (QoS) requirements are met at the same time while considering interference degree. Furthermore, a novel conflict graph model is constructed to represent the conflicts among transmission links. This model is then transformed into a conflict hypergraph structure, enabling a more comprehensive and accurate depiction of complex interference relationships within the network. We convert the combinatorial optimization problem for radio resource allocation into a Markov decision process (MDP) model. For IoV with dense interference, we employ a federated averaging-deep Q-Network (FedAvg-DQN) algorithm to enhance the transmission rate and overall network throughput. Consequently, the experimental findings demonstrate that this method can significantly enhance the anti-interference capacity of the system in a densely interfered Internet of Vehicles (IoV) environment. Specifically, compared to the benchmark algorithm, network throughput is on average increased by 28.17<span><math><mtext>%</mtext></math></span>, and energy efficiency is on average improved by 24.85<span><math><mtext>%</mtext></math></span>. These improvements effectively accelerate the information transmission rate and guarantee the reliability of the information transmission. These findings validate the efficacy and practicality of the developed approach, demonstrating its potential to enhance the performance of IoV systems under challenging interference conditions.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"269 ","pages":"Article 111402"},"PeriodicalIF":4.4,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144254520","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":"Metaversal intelligence: Unifying human-AI interactions in human-in-the-loop AIB-Metaverse","authors":"Minoo Soltanshahi,&nbsp;Martin Maier","doi":"10.1016/j.comnet.2025.111425","DOIUrl":"10.1016/j.comnet.2025.111425","url":null,"abstract":"<div><div>The convergence of 6G and Web 3 technologies with the Metaverse necessitates user-centric, intelligent solutions to address challenges of adaptability, scalability, and efficiency in decentralized systems. This paper introduces the Human-AI Blockchain (HAIB) Metaverse architecture and the Intelligent Smart Contract (InSC) framework, which integrate Reinforcement Learning (RL) and extended stigmergy to optimize multi-agent interactions and enable dynamic, self-executing smart contracts. The HAIB architecture enhances human-AI collaboration through a human-in-the-loop approach, facilitating adaptive decision-making and efficient resource allocation. Serving as the cognitive core, the InSC framework combines dynamic environmental feedback with real-time data analytics to enable intelligent decision-making. Experimental evaluations demonstrate that the proposed framework reduces gas consumption by 32–33% while increasing cumulative rewards. These findings highlight the significant potential of intelligent smart contracts in advancing decentralized intelligence within the Metaverse.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"269 ","pages":"Article 111425"},"PeriodicalIF":4.4,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144230255","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":"DBSRP-ML: Dynamic blockchain sharding reconfiguration protocol based on multi-label","authors":"Hui Dai ,&nbsp;Lingyun Yuan ,&nbsp;Jiaying Wu ,&nbsp;Han Chen ,&nbsp;Haochen Bao","doi":"10.1016/j.comnet.2025.111401","DOIUrl":"10.1016/j.comnet.2025.111401","url":null,"abstract":"<div><div>Sharding provides an effective approach to improving the scalability of blockchain. To address the problems of cross-shard transaction proliferation, transaction latency increase, and load imbalance caused by the improper allocation of shardings, we propose a dynamic blockchain sharding reconfiguration protocol based on multi-label (DBSRP-ML). First, the account state sharding model based on the label graph network is constructed, which achieves the fine-grained correlation analysis and efficient management of the topology of the account transaction network by decoupling and reconstructing the account states. Second, we propose the account label multiplication mechanism to enable efficient mapping between accounts and shards. Third, we propose the multi-label account sharding partition algorithm (MLASPA) to optimize the sharding configuration according to the optimal sharding belonging coefficient of accounts, which enhances the degree of accurate sharding matching, thus minimizing the number of cross-shard transactions. Experimental results show that compared with other sharding protocols, DBSRP-ML improves the system throughput by 50% and up to 268%, reduces the latency by 72.8% to 88%, reduces the cross-shard ratio by up to 70%, and displays excellent workload balancing. The proposed protocol improves the scalability of the blockchain, thereby optimizing its overall performance.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"269 ","pages":"Article 111401"},"PeriodicalIF":4.4,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144240464","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":"An effective data dissemination method in WAIC networks using enhanced CTC","authors":"Tao Cheng ,&nbsp;Shining Li ,&nbsp;Weiwei Dang ,&nbsp;Yan Pan","doi":"10.1016/j.comnet.2025.111399","DOIUrl":"10.1016/j.comnet.2025.111399","url":null,"abstract":"<div><div>By introducing wireless technology into aircraft systems, Wireless Avionics Intra-Communication (WAIC) networks aim to enhance the efficiency and flexibility of onboard operations. Rather than replacing traditional wired systems, WAIC enables the seamless coexistence of heterogeneous wireless nodes to address diverse application requirements, with a particular focus on efficient network management and time-critical data transmission. This paper proposes an effective data dissemination method leveraging cross-technology communication (CTC) to achieve direct data delivery from high-speed gateways to multi-hop field nodes in coexisting hybrid WAIC subnets. The proposed model relies on the efficiency of CTC and its effective transmission capability in multi-hop networks. We introduce an optimized signal emulation method inspired by power-domain non-orthogonal multiple access (PD-NOMA) in 5G, which superimposes multiple signals to improve emulation accuracy from 802.11g to 802.15.4. We also develop a multi-hop CTC link quality evaluation model, providing a comprehensive analysis of reception performance across varying hop counts. Based on these advancements, we design a coordinated data dissemination method that exploits the high-gain propagation and direct communication capabilities of CTC, significantly reducing the latency of multi-hop relaying. Extensive experimental results demonstrate the effectiveness of the proposed hybrid model. In a typical setting, it achieves a 3.18<span><math><mo>×</mo></math></span> reduction in the maximum network dissemination delay and a 2.83<span><math><mo>×</mo></math></span> decrease in the average dissemination delay, compared to traditional hop-by-hop transmissions.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"269 ","pages":"Article 111399"},"PeriodicalIF":4.4,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144264084","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":"Cache-INT: In-network caching-enabled In-band Network Telemetry","authors":"Penghui Zhang ,&nbsp;Hua Zhang ,&nbsp;Yuqi Dai ,&nbsp;Yibo Pi ,&nbsp;Jingyu Wang ,&nbsp;Jianxin Liao","doi":"10.1016/j.comnet.2025.111404","DOIUrl":"10.1016/j.comnet.2025.111404","url":null,"abstract":"<div><div>Due to its powerful monitoring capabilities, In-band Network Telemetry (INT) technology has become essential to modern network management. However, a key challenge lies in reducing the redundant telemetry data transmission from the data plane to the controller, a vital process for real-time network status monitoring and swift decision-making. This challenge, if not properly addressed, could significantly waste network bandwidth and increase the controller’s response time.</div><div>In this paper, we present Cache-INT, a highly efficient in-network caching-enabled INT system. Specifically, Cache-INT stores and reuses network information collected by probes through a well-designed cache strategy, and uses the incremental transmission technique to reduce the volume of data transmitted. Furthermore, an INT system’s performance depends heavily on probe path planning, which determines both the network coverage and control overhead. We optimize these factors in two scenarios: one with fixed cache-enabled routers, and the other allowing for flexible selection of cache-enabled routers. For the former, we use a Deep Reinforcement Learning (DRL)-based algorithm to minimize control overhead and design a mask function to speed up training. For the latter, we combine the simulated annealing algorithm and the DRL model to find a suitable cache-enabled router upgrade scheme, and then plan the probe paths accordingly.</div><div>The experimental results show that, compared to existing INT systems, Cache-INT reduces the data transmission volume from the data plane to the controller by at least 40% while achieving a 30% reduction in control overhead.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"269 ","pages":"Article 111404"},"PeriodicalIF":4.4,"publicationDate":"2025-06-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144254517","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}