Computer Networks最新文献

{"title":"D2BFT: A dual Byzantine fault tolerance approach for multi-agent drone surveillance with deep reinforcement learning","authors":"Viswesh Nanapu,&nbsp;Gopi Banavathu,&nbsp;Srinivasa Desikan KE","doi":"10.1016/j.comnet.2025.111750","DOIUrl":"10.1016/j.comnet.2025.111750","url":null,"abstract":"<div><div>The Dual Byzantine Fault Tolerance (D2BFT) framework integrates Practical Byzantine Fault Tolerance (PBFT) and Delegated Byzantine Fault Tolerance (DBFT) within a Multi-Agent Reinforcement Learning Proximal Policy Optimization (MARLPPO) environment to secure autonomous drone swarms for urban surveillance. Under a two-phase protocol, a subset of delegated validators conducts a fast PBFT-style consensus that a secondary group of validators then verifies. Deployed on Unity, D2BFT resists up to 40% malicious agents and provides a 20% improvement in the average consensus latency (0.60 s compared to 0.75 s for PBFT) with throughput of more than 30 transactions per second. Through realistic simulation of fault injections, Invert, Randomize, Silent, and Conflicting D2BFT ensure strong agreement with negligible overhead in the presence of high mobility and spotty connectivity. These findings validate the capacity of D2BFT to maintain resilience and efficiency, providing a scalable approach to fault-ridden real-time drone networks.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111750"},"PeriodicalIF":4.6,"publicationDate":"2025-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364296","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":"Symbiotic agents: A novel paradigm for trustworthy AGI-driven networks","authors":"Ilias Chatzistefanidis ,&nbsp;Navid Nikaein","doi":"10.1016/j.comnet.2025.111749","DOIUrl":"10.1016/j.comnet.2025.111749","url":null,"abstract":"<div><div>Large Language Model (LLM)-based autonomous agents are expected to play a vital role in the evolution of 6G networks, by empowering real-time decision-making related to management and service provisioning to end-users. This shift facilitates the transition from a specialized intelligence approach, where artificial intelligence (AI) algorithms handle isolated tasks, to artificial general intelligence (AGI)-driven networks, where agents possess broader reasoning capabilities and can manage diverse network functions. In this paper, we introduce a novel agentic paradigm that combines LLMs with real-time optimization algorithms towards Trustworthy AI, defined as <em>symbiotic agents</em>. Optimizers at the LLM’s input-level provide bounded uncertainty steering for numerically precise tasks, whereas output-level optimizers supervised by the LLM enable adaptive real-time control. We design and implement two novel agent types including: (i) Radio Access Network (RAN) optimizers, and (ii) multi-agent negotiators for Service-Level Agreements (SLAs). We further propose an end-to-end architecture for AGI-driven networks and evaluate it on a 5G testbed capturing channel fluctuations from moving vehicles. Results show that symbiotic agents reduce decision errors fivefold compared to standalone LLM-based agents, while smaller language models (SLM) achieve similar accuracy with a 99.9 % reduction in Graphical Processing Unit (GPU) resource overhead and in near-real-time (near-RT) loops of <span><math><mrow><mn>82</mn><mspace></mspace><mi>m</mi><mi>s</mi></mrow></math></span>. A multi-agent demonstration for collaborative RAN on the real-world testbed highlights significant flexibility in service-level agreement and resource allocation, reducing RAN over-utilization by approximately 44 %. Drawing on our findings and open-source implementations, we introduce the symbiotic paradigm as the foundation for next-generation, AGI-driven networks-systems designed to remain adaptable, efficient, and trustworthy even as LLMs advance. A live demo is presented here <span><span>https://www.youtube.com/watch?v=WQv61z1deXs&amp;ab_channel=BubbleRAN</span><svg><path></path></svg></span></div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111749"},"PeriodicalIF":4.6,"publicationDate":"2025-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145363853","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":"ProxyCorr: robust traffic correlation attacks via mixed spatio-temporal analysis in encrypted proxy networks","authors":"Mengyan Liu ,&nbsp;Gaopeng Gou ,&nbsp;Gang Xiong ,&nbsp;Junzheng Shi ,&nbsp;Zhong Guan ,&nbsp;Hanwen Miao ,&nbsp;Chen Chen","doi":"10.1016/j.comnet.2025.111763","DOIUrl":"10.1016/j.comnet.2025.111763","url":null,"abstract":"<div><div>End-to-end traffic correlation attacks, aiming to deanonymize users in anonymous communication systems, have achieved significant advancements through deep learning. However, existing methods suffer significant limitations in encrypted proxy networks, especially when obfuscation techniques are employed. To address these limitations, we propose ProxyCorr, a model that achieves accurate traffic correlation across diverse encrypted proxy protocols. Specifically, we model network flows as state sequences and derive temporal dependencies through state transition analysis. Additionally, to capture spatial similarity patterns, we model flows as 2D spatial trajectories, with Gaussian filtering applied to mitigate jitter. Further, we design a self-attention-based correlation learning module to extract inherent spatio-temporal features, which are then adaptively integrated to enhance flow correlation performance. Extensive experiments on multiple encrypted-proxy datasets demonstrate that ProxyCorr outperforms state-of-the-art methods in effectiveness while maintaining robustness against obfuscation techniques, including packet size padding and multiplexing.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111763"},"PeriodicalIF":4.6,"publicationDate":"2025-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145326128","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":"LSTM-TRPS: Trajectory reconstruction protection strategy based on semantic information encoding","authors":"Jing Zhang ,&nbsp;Tongxin Li ,&nbsp;Haoze Hu ,&nbsp;Huaxiong Liao ,&nbsp;Xiu-Cai Ye","doi":"10.1016/j.comnet.2025.111768","DOIUrl":"10.1016/j.comnet.2025.111768","url":null,"abstract":"<div><div>Mobility data from IoT devices are vulnerable to attacks, and existing privacy methods often fail to defend against background knowledge adversaries. To mitigate this risk, a Long Short Term Memory based Trajectory Reconstruction Protection Strategy (LSTM-TRPS) is proposed to generate utility-preserving but unidentifiable synthetic trajectories. LSTM-TRPS is designed as a post-processing defense, maintaining essential mobility patterns for analytics while blocking reconstruction attempts. It consists of four modules: (1) Point of interest (POI) Semantic Annotation (PSA) for geo-temporal labeling; (2) Hasse Diagram-based Semantic Encoding (HDSE) to preserve hierarchical semantics; (3) Feature Embedding and Adaptive Matrix Combination (FEAMC) to fuse spatial temporal features; and (4) a Bi-LSTM generator to produce robust trajectories. Experiments on real-world mobility datasets show that LSTM-TRPS reduces Hausdorff distance by 12.7 %, improves temporal alignment by 20 %, and lowers privacy leakage by over 50 % under strict privacy budgets. It also achieves over 90 % POI retention and strong generalization across datasets, making it well suited for privacy-preserving trajectory publishing in IoT and smart mobility scenarios.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111768"},"PeriodicalIF":4.6,"publicationDate":"2025-10-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145326129","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":"Truthful reverse auction-based incentive mechanisms for task offloading in mobile edge computing","authors":"Xueyi Wang ,&nbsp;Yi Ma ,&nbsp;Zhihao Qin ,&nbsp;Jian Xu ,&nbsp;Jianzhe Zhao ,&nbsp;Rongfei Zeng ,&nbsp;Yang Song ,&nbsp;Qiang He","doi":"10.1016/j.comnet.2025.111767","DOIUrl":"10.1016/j.comnet.2025.111767","url":null,"abstract":"<div><div>Offloading computation-intensive tasks of mobile devices (MDs) to the nearby base stations (BSs) equipped with edge servers has been a promising approach to effectively alleviate the problem of insufficient computation capacity of MDs. Nevertheless, task offloading needs to consume numerous resources (e.g., computation and communication resources, etc.). Assumed that the owners of BSs are selfish and rational, they will be reluctant to take part in the task offloading without acquiring suitable economic reimbursement. Therefore, it is necessary to design an effective incentive mechanism to motivate BSs to engage in the process of task offloading. In this paper, we present a truthful reverse auction-based incentive mechanism (TRAIM) with explicit consideration of the locality constraint, overlapped coverage and capacity constraint to offload tasks remotely. To be specific, we first devise the <em>MD assignment</em> based on the dynamic programming approach to acquire the set of optimal BS-MD associations. Next, we devise the <em>winning BS selection</em> using the greedy method to decide the set of winning BSs. Then, we compute the charge of each winning BS according to its critical value, aiming to prevent the strategic behaviors. Strict theoretical analysis demonstrates our proposed TRAIM is truthful, individually rational and computationally efficient. Extensive simulations also demonstrate the effectiveness of our proposed TRAIM.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111767"},"PeriodicalIF":4.6,"publicationDate":"2025-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145326126","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":"Global consistent graph convolutional network for amplifying piezoelectric microelectromechanical sensor performance in fluid-dynamic environments","authors":"S.K. Mydhili ,&nbsp;Elangovan Muniyandy ,&nbsp;Rajeshkannan S ,&nbsp;T.R. Vijaya Lakshmi","doi":"10.1016/j.comnet.2025.111772","DOIUrl":"10.1016/j.comnet.2025.111772","url":null,"abstract":"<div><div>Microelectromechanical System (MEMS) sensors composed of a bluff body and a polyvinylidene fluoride (PVDF) flexible piezoelectric flag, are highly sensitive to variations in fluid dynamics. However, the sensing capability of these systems is limited by factors such as complex vortex formation and inaccurate turbulence classification, especially under varying fluid speeds and bluff body geometries. To address these challenges, the Global Consistent Graph Convolutional Network for Amplifying Sensing Capability of Piezoelectric Microelectromechanical System Sensors in a Fluid-Dynamic System (GCGCN-PMEMS-FDS) is proposed for accurate turbulence classification. The system first inputs wind or fluid speed, inducing mechanical vibrations in piezoelectric flag, which displaces charge and generates voltage signals. These signals are then processed using the Morlet Scattering Transform (MST) to extract wind speed features, such as higher and lower wind speeds. The extracted features are fed into GCGCN to classify turbulence levels into low, moderate, and high. To validate the proposed method, experiments were conducted using various wind speeds and bluff body designs in a wind tunnel. Implemented in Python, the GCGCN-PMEMS-FDS approach demonstrated superior performance compared to existing methods, achieving higher accuracy of 99.92 % in turbulence classification, low computation time of 93 s compared with existing methods. These results highlight the effectiveness of the GCGCN-PMEMS-FDS method in enhancing sensing capabilities of MEMS sensors in fluid-dynamic systems.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111772"},"PeriodicalIF":4.6,"publicationDate":"2025-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145363851","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":"Deadline-aware load balancing for coflow in datacenter networks","authors":"Zhi Gong ,&nbsp;Zhichen Wang ,&nbsp;Jinbin Hu ,&nbsp;Jin Wang ,&nbsp;Fayez Alqahtani ,&nbsp;Amr Tolba ,&nbsp;Kai Jin","doi":"10.1016/j.comnet.2025.111765","DOIUrl":"10.1016/j.comnet.2025.111765","url":null,"abstract":"<div><div>To accurately capture the characteristics of collective data transmission in modern distributed data parallel applications, the abstraction of Coflow is introduced. Notably, a significant number of Coflows in time-sensitive applications have strict deadlines, which are critical for meeting the application’s timeliness and performance requirements. However, existing Coflow transmission mechanisms fail to fine-grain the classification of Coflows, and current load balancing strategies do not holistically consider the transmission characteristics of Coflows. This inevitably increases Coflow completion time (CCT) and raises the proportion of Coflows missing their deadlines. To address this issue, we propose DLB. Specifically, DLB classifies Coflows into three transmission levels. Coflows with deadlines that can be completed within their respective deadlines are categorized as the first transmission level. Furthermore, DLB prioritizes routing paths based on the queue length at routing ports and the queue growth rate of transmission paths, ensuring that first-level Coflows are directed to the highest-priority routing ports. The evaluation results show that DLB reduces the miss rate up to 33%, 28%, 23%, 15% and 6% over ECMP, LetFlow, Varys, Con-Myopic and JFSR at a load of 0.7 under Web Search workload, respectively.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111765"},"PeriodicalIF":4.6,"publicationDate":"2025-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145326127","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":"Defending against smishing attacks: State-of-the-art techniques, challenges, limitations, and future directions","authors":"Mosiur Rahaman ,&nbsp;Nicko Cajes ,&nbsp;Brij B. Gupta ,&nbsp;Kwok Tai Chui ,&nbsp;Nadia Nedjah","doi":"10.1016/j.comnet.2025.111758","DOIUrl":"10.1016/j.comnet.2025.111758","url":null,"abstract":"<div><div>Smishing is a hostile operation by attackers to get sensitive information using misrepresentation. Because autonomous anti-phishing systems are not very accurate, they may miss smishing crimes that are facilitated by thoughtful phishing SMS, or short message Smishing attacks have become more common over the past few years because of the widespread adoption of these user-friendly and useful devices. The goal of this literature review is to investigate the approaches and strategies employed in smishing attacks by utilizing categorization methods. This article outlines the current systems that employ deep learning and machine learning methods, along with their advantages, disadvantages, and restrictions. To effectively prevent spam messages rather than detect them, this study discusses the potential for future advancements in processing natural languages. We figured out about psychological issues, argumentation and need, bias against confirmation, and ignorance, among other smishing-related insights, problems, and future research directions. These findings suggest that understanding the cognitive process and its workings is essential to developing solutions for smishing attack detection and mitigation.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111758"},"PeriodicalIF":4.6,"publicationDate":"2025-10-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145326508","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":"PrivHFL: A privacy-preserving scheme for hierarchical federated learning","authors":"Bayan Alzahrani,&nbsp;Dejun Yang","doi":"10.1016/j.comnet.2025.111764","DOIUrl":"10.1016/j.comnet.2025.111764","url":null,"abstract":"<div><div>In the Internet of Things (IoT) field, where interconnected devices generate sensitive data, ensuring privacy is a major challenge. Federated Learning (FL) addresses this by allowing devices to collaboratively train a model without sharing their local data, improving privacy in IoT systems. While the traditional two-layer FL framework is commonly used, adopting a hierarchical client-edge-cloud architecture can significantly accelerate model training, especially in resource-constrained IoT networks. Hierarchical Federated Learning (HFL) offers significant advantages, yet concerns persist regarding potential privacy breaches from analyzing client or edge server data. To address these concerns, we propose PrivHFL, a privacy-preserving solution for HFL that leverages threshold homomorphic encryption. Security and performance analyses demonstrate that the proposed scheme is scalable, supporting larger FL scenarios, including diverse IoT environments, while ensuring data privacy. PrivHFL is resilient to collusion among nearly half of the clients and effectively handles client dropouts. Our approach achieves high accuracy in IID and non-IID scenarios, as demonstrated using the MNIST, CIFAR-10, and CIFAR-100 datasets. Additionally, we show that the added encryption overhead is reasonable, making our solution feasible for real-world IoT applications.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111764"},"PeriodicalIF":4.6,"publicationDate":"2025-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145326546","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":"PowerNetMax: A DRL-GNN framework for IRS-Assisted IOT network optimization","authors":"Muhammad Farhan ,&nbsp;Lei Wang ,&nbsp;Nadir Shah ,&nbsp;Gabriel-Miro Muntean ,&nbsp;Awais Bin Asif ,&nbsp;Houbing Herbert Song","doi":"10.1016/j.comnet.2025.111760","DOIUrl":"10.1016/j.comnet.2025.111760","url":null,"abstract":"<div><div>Intelligent Reflecting Surfaces (IRS) have recently emerged as a cutting-edge technology in 6G Internet of Things (IoT) communications, offering substantial connectivity enhancements, particularly in remote, high-mobility, or obstacle-prone environments. This paper proposes PowerNetMax, an innovative framework designed to improve overall network connectivity, reliability, and energy efficiency in IRS-assisted IoT communication systems. PowerNetMax leverages a comprehensive set of network parameters and integrates the strengths of Deep Reinforcement Learning (DRL) and Graph Neural Networks (GNN) to enable intelligent and adaptive optimization. Through extensive experimentation, PowerNetMax demonstrates up to 5–20 % higher received power, 50 % faster convergence, and 20 % higher throughput under mobility conditions compared to state-of-the-art GNN-based and heuristic solutions. Extensive simulation results confirm that PowerNetMax achieves superior adaptability and robustness, highlighting its effectiveness for future IRS-assisted IoT networks.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111760"},"PeriodicalIF":4.6,"publicationDate":"2025-10-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145326509","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}