Computer Networks最新文献

{"title":"An improved metric-active learning approach for few labeled radio frequency fingerprinting","authors":"Chuan He ,&nbsp;Qingchun Meng ,&nbsp;Yao Chen ,&nbsp;Tao Zhang ,&nbsp;Guyue Li","doi":"10.1016/j.comnet.2025.111794","DOIUrl":"10.1016/j.comnet.2025.111794","url":null,"abstract":"<div><div>Radio frequency fingerprinting (RFF) is an effective non-cryptographic method for physical-layer authentication. Current deep learning (DL) approaches have achieved strong results in RFF identification but typically require large annotated datasets, making data collection and labeling costly. To address this, we propose a novel Active Learning (AL) framework that builds a robust classifier with minimal labeled data through incremental learning. Our framework improves AL in two ways: (1) using a Siamese-based metric learning model to capture discriminative features from data-pair similarities, and (2) adopting a cost-effective sample selection strategy that reduces manual labeling while enhancing accuracy. Unlike methods that focus only on low-confidence samples, our approach also leverages high-confidence samples from the unlabeled pool, assigning them pseudo-labels to expand the training set. Experiments on Laboratory LoRa devices show that the framework achieves superior performance with fewer labeled samples.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111794"},"PeriodicalIF":4.6,"publicationDate":"2025-10-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145363855","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":"Flow persona: A QoS Flow Rule Scheme based on Deep Learning in SDN-IoT","authors":"Hao She ,&nbsp;Lixing Yan ,&nbsp;Xin An ,&nbsp;Chuanfeng Mao ,&nbsp;Yongan Guo","doi":"10.1016/j.comnet.2025.111784","DOIUrl":"10.1016/j.comnet.2025.111784","url":null,"abstract":"<div><div>With the proliferation of Internet of Things (IoT) devices and increasing network flow, traditional network architectures struggle to manage complex flow and meet evolving Quality of Service (QoS) requirements. These architectures lack flexibility in resource allocation and optimization, limiting their support for diverse IoT applications. To address these issues, we propose a QoS Flow Rule Scheme based on Deep Learning in Software Defined Networking-IoT (SDN-IoT) called Flow Persona. This scheme integrates user personas and QoS requirements, employs an ARIMA model for traffic prediction, and leverages a Convolutional Neural Network (CNN) optimized by Adaptive Particle Swarm Optimization (APSO) for flow classification. Simulation results show that flow persona improves QoS flow classification accuracy by about 4.6% over traditional and existing algorithms. It also significantly enhances precision, recall, and F-score, while improving QoS routing efficiency and reducing network delay.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111784"},"PeriodicalIF":4.6,"publicationDate":"2025-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145363852","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":"LRCC: Long-haul RDMA congestion control for cross-datacenter networks","authors":"Dingyu Yan ,&nbsp;Yaping Liu ,&nbsp;Shuo Zhang ,&nbsp;Mingguang Xu ,&nbsp;Zhikai Yang ,&nbsp;Binxing Fang","doi":"10.1016/j.comnet.2025.111756","DOIUrl":"10.1016/j.comnet.2025.111756","url":null,"abstract":"<div><div>With the widespread deployment of applications such as cloud storage and distributed model training, Remote Direct Memory Access (RDMA) is increasingly applied to cross-datacenter networks. These networks typically consist of multiple regional datacenters interconnected by dedicated long-haul optical fiber and Data Center Interconnect (DCI) switches. However, existing RDMA congestion control mechanisms face significant challenges in cross-datacenter networks. Firstly, the long control loops struggle to effectively suppress line-rate bursts of cross-domain RDMA traffic, leading to persistent queues that degrade overall network performance. Secondly, the heterogeneous Round-Trip Time (RTT) characteristics between cross-domain and intra-datacenter traffic disrupt the convergence and fairness guarantees of conventional methods, further exacerbating cross-domain congestion issues. In this paper, we propose a switch-driven Long-haul RDMA Congestion Control (LRCC). LRCC utilizes near-source switches to generate congestion notification packets, effectively shortening the long control loops. Furthermore, LRCC implements a precise fair-rate computation mechanism on the switches and an adaptive rate-increase strategy on the host. These mechanisms mitigate cross-domain congestion caused by hybrid traffic while ensuring high throughput for long-haul flows. We implemented a prototype system of LRCC on programmable switches and 400Gbps FPGA NICs. Testbed experiments show that, compared with the NVIDIA CX7, LRCC reduces tail latency by 11 %-16 % in short-distance congestion scenarios and by 45 %-49 % in a 640 km long-distance scenario. Large-scale simulations further demonstrate that in the cross-datacenter networks, LRCC outperforms existing solutions, reducing the average Flow Completion Time (FCT) by up to 67.2 %, 94 % and 48.4 %, respectively, compared to DCQCN, HPCC and BiCC.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111756"},"PeriodicalIF":4.6,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145363849","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":"IoT device identification method based on transformer and clustering","authors":"Litong Deng ,&nbsp;Dinglin Gu ,&nbsp;Zhi Lin","doi":"10.1016/j.comnet.2025.111791","DOIUrl":"10.1016/j.comnet.2025.111791","url":null,"abstract":"<div><div>With the rapid proliferation of Internet of Things (IoT) technologies, mitigating unauthorized device intrusions and impersonation attacks has become a critical security challenge. Device identification plays a crucial role in detecting anomalous behaviors, thereby enhancing security during device operation. However, existing identification methods predominantly rely on manually crafted feature engineering, which necessitates extensive domain knowledge and involves a time-consuming feature selection process. This not only increases computational overhead but also risks omitting essential information, thereby limiting identification performance. To address these challenges, this paper proposes a sample construction method that converts network traffic into multibyte token sequences, utilizes the Transformer architecture to model both the temporal and contextual relationships of raw traffic packets. This approach eliminates the need for complex feature engineering and enables efficient sample generation from just one minute of network traffic, facilitating accurate and efficient IoT device identification. To tackle the open-set identification problem and enhance security management during device access, this study extends the end-to-end identification framework by integrating metric learning with HDBSCAN clustering to generate distinctive device fingerprints. This method not only effectively classifies known devices but also reliably detects previously unseen devices. Experimental results on two public datasets, UNSW and Yourthings, demonstrate that the proposed method achieves superior performance, attaining accuracy rates of 99.89 % and 99.68 %, respectively. Furthermore, it outperforms existing approaches in terms of recognition accuracy, generalization capability, and scalability.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111791"},"PeriodicalIF":4.6,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145363854","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":"ELVIS: eBPF-based extensions of linux hosts for using virtual network functions with service function chaining and in-band network telemetry","authors":"Marco Haeberle ,&nbsp;Benjamin Steinert ,&nbsp;Michael Weiss ,&nbsp;Michael Menth","doi":"10.1016/j.comnet.2025.111773","DOIUrl":"10.1016/j.comnet.2025.111773","url":null,"abstract":"<div><div>Service function chaining (SFC) is a technology that enables dynamic steering of packets to processing nodes, also called Service Functions (SFs), e.g., firewalls, IDSs, or NAT gateways. An SFC classifier located at the border of SFC-enabled domains encodes information about the order of SFs, the Service Function Chain, into packet headers. This encoded information, called SFC encapsulation, is used by the network to steer packets to the respective SFs. Because most existing SFs do not support SFC encapsulations natively, a proxy is required to make them SFC-compatible. However, simple proxies operate in a stateless manner and are not capable of preserving per-packet metadata. In this paper, we propose a dynamic SFC proxy that is capable of preserving metadata by caching the SFC encapsulation while a packet is processed by a SF. In addition, the proxy implements In-band Network Telemetry (INT) to support monitoring and debugging related to SFs. INT is a network monitoring framework that adds packet-specific metrics to the header stack of a packet. Although INT has been standardized with a focus on network switches and routers, a proxy-based implementation for SFs extends its usability in an SFC scenario. We present concept, use cases, and an eBPF-based implementation of the INT-enabled caching SFC proxy. Finally, we evaluate the performance of a prototype.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111773"},"PeriodicalIF":4.6,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364297","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":"Edge AI-based self-learning technique for mitigating DDoS attacks in WSN","authors":"Saqib Hussain ,&nbsp;Jingsha He ,&nbsp;Nafei Zhu ,&nbsp;Fahad Razaque Mughal ,&nbsp;Sadique Ahmad ,&nbsp;Muhammad Iftikhar Hussain ,&nbsp;Zulfiqar Ali Zardari","doi":"10.1016/j.comnet.2025.111769","DOIUrl":"10.1016/j.comnet.2025.111769","url":null,"abstract":"<div><div>Wireless Sensor Networks (WSNs) are increasingly deployed in critical applications but remain highly susceptible to Distributed Denial of Service (DDoS) attacks. The constrained nature of WSNs-limited energy, memory, and processing power-renders traditional centralized intrusion detection systems inefficient due to high latency, bandwidth consumption, and privacy risks. This research presents a lightweight, edge AI-based self-learning framework that leverages federated deep learning to detect and mitigate DDoS attacks in WSNs without sharing raw sensor data. The proposed model enables local model training at individual nodes and aggregates updates securely at a central server, preserving data privacy while reducing network overhead. A multi-layer deep learning architecture is utilized to enhance anomaly detection capability across heterogeneous devices. Experimental results demonstrate the effectiveness of the approach, achieving 99.2 % accuracy, 96 % precision, 94 % recall, and improved corresponding F1-scores on benchmark datasets. Compared to traditional centralized models, the framework reduces latency by 31.7 % and communication overhead by 28 %, while maintaining high detection performance even under Non-IID data distributions. Additionally, it shows resilience to class imbalance and scalability across distributed nodes. These results confirm the framework’s suitability for real-time, privacy-preserving intrusion detection in modern WSNs. Future work will extend this system with adaptive client selection, adversarial robustness techniques, and deployment on live WSN hardware to validate real-time performance and security under evolving threat conditions.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111769"},"PeriodicalIF":4.6,"publicationDate":"2025-10-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145363850","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":"Maximizing end-to-end TCP connection’s aggregated throughput in connected vehicle networks","authors":"M. Joseph Auxilius Jude,&nbsp;M. Shivaranjani","doi":"10.1016/j.comnet.2025.111785","DOIUrl":"10.1016/j.comnet.2025.111785","url":null,"abstract":"<div><div>Transmission control protocol (TCP) handles a considerable proportion of Internet traffic in connected vehicular networks (CVNs) and encounters a persistent issue with the faulty activation of the rate regulation mechanism during random packet losses or latency fluctuations as well as false timeouts during delayed acknowledgments (ACKs). An enhanced Cubic (<em>e-Cubic</em>) TCP is proposed in this work, which employs enhanced congestion control (<em>e-CC</em>) and enhanced retransmission timeout (<em>e-RTO</em>) algorithms to improve TCP's traffic performance in CVN links. The <em>e-CC</em> algorithm facilitates the accelerated cubic growth function by introducing a new window increment pattern and adjusting the appropriate scaling and elapsed time parameters. Furthermore, <em>e-CC</em> implements a dual verification mechanism to identify congestion and regulate unnecessary rate regulation caused by losses in wireless transmission. The <em>e-RTO</em> algorithm computes and sets the optimal timeout duration to receive delayed ACKs during latency spikes in wireless links, thereby reducing the frequency of false timeouts. The simulation results demonstrate that <em>e-Cubic</em> reduces packet delays and packet loss while increasing the total throughput and delivery rate of TCP traffic in CVN links.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111785"},"PeriodicalIF":4.6,"publicationDate":"2025-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145326505","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":"Spectral reinforcement learning based dynamic routing for unmanned aerial vehicle (UAV) networks","authors":"Saif ullah ,&nbsp;Khalid Hussain ,&nbsp;Muhammad Faheem ,&nbsp;Nisar Ahmed Memon","doi":"10.1016/j.comnet.2025.111787","DOIUrl":"10.1016/j.comnet.2025.111787","url":null,"abstract":"<div><div>Unmanned Aerial Vehicles (UAVs) have received a lot of interest for their prospective uses in various types of disciplines, including communication, disaster management, surveillance, and military applications. UAV ad-hoc networks enable UAVs to interact wirelessly without a permanent infrastructure, making them suited for many circumstances. Conventional methods require predefining the number of clusters, which can lead to inaccurate results, and existing schemes focus on distance as the key parameter while neglecting UAV connectivity; additionally, traditional algorithms struggle with complex UAV network structures due to varying distances, obstacles, and dynamic configurations, making them unable to adapt to frequent changes in connectivity, signal strength, and network topology. This study proposes a framework that integrates spectral clustering and reinforcement learning to optimize the performance of UAV ad hoc networks. Spectral clustering groups UAVs with similar communication characteristics, such as signal strength and geographic location. Reinforcement learning is then used to optimize the path UAVs take within each clustered group, leading to further improvements in network performance. Our approach effectively adapts to changes in network topology and communication patterns, allowing for optimal performance even in dynamic environments. Experimental results demonstrate the effectiveness of our strategy, achieving a Packet Delivery Ratio (PDR) improvement of approximately 18.42% over k-means routing at high mobility scenarios, with an end-to-end delay reduction of around 40% compared to traditional methods. Additionally, the Network Routing Load (NRL) of our proposed scheme remains consistently below 18%, indicating enhanced efficiency compared to existing protocols, which can reach NRL values of up to 35%. Our approach optimizes the communication efficiency of UAV ad-hoc networks by adopting an optimal route policy, resulting in reduced end-to-end delay and improved packet delivery ratio. The proposed framework offers several advantages over existing methods, including adaptability to changes in network topology and communication patterns, efficient communication, and optimal routing decisions.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111787"},"PeriodicalIF":4.6,"publicationDate":"2025-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145326500","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":"Switch bypass: End-host cloud networking revisited","authors":"Antonio Le Caldare ,&nbsp;Luigi Leonardi ,&nbsp;Sebastiano Miano ,&nbsp;Gregorio Procissi ,&nbsp;Gianni Antichi ,&nbsp;Giuseppe Lettieri","doi":"10.1016/j.comnet.2025.111725","DOIUrl":"10.1016/j.comnet.2025.111725","url":null,"abstract":"<div><div>Virtual switches are one of the most important building blocks in public cloud network stacks as they apply high-level policies to traffic enabling communication between virtual machines (VMs) and the rest of the world. The problem is that virtual switches need CPU cores to process packets and the more cores assigned to them, the less are available to VMs that are rented to customers and hence generate revenue.</div><div>With this paper, we show that it is potentially possible to find a sweet-spot between performance and costs. The insight is that applications running on VMs are not always using 100 % of their CPU processing power: we use this to design <em>switch bypass</em>, a new technique that allow virtual switches to opportunistically offload part of their processing to the virtual NIC drivers associated with guest VMs.</div><div>Using packet classification as use-case, we show that with <em>switch bypass</em> we obtain a performance boost up to 80 % without the need of additional core processing power.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111725"},"PeriodicalIF":4.6,"publicationDate":"2025-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145326507","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":"EE-AIRP: An AI-enhanced energy-efficient routing protocol for IoT-enabled WSNs","authors":"Nguyen Duy Tan,&nbsp;Nguyen Minh Quy,&nbsp;Van-Hau Nguyen","doi":"10.1016/j.comnet.2025.111770","DOIUrl":"10.1016/j.comnet.2025.111770","url":null,"abstract":"<div><div>Wireless sensor networks (WSNs) have become essential components in Internet of Things (IoT) applications for data collection and processing. The constrained energy resources and limited computational capacities inherent in sensor nodes impose significant limitations on the operational lifespan of WSNs, thereby constituting a persistent and critical research challenge in the field. This paper proposes an Energy-Efficient Artificial Intelligence-based Routing Protocol (EE-AIRP) to extend network lifespan in WSN-based IoT applications. The proposed methodology offers three principal innovations that collectively advance the current state of research: (1) network zone partitioning using the DBSCAN machine learning algorithm to form approximately balanced clusters based on node distribution density, (2) intelligent cluster head (CH) selection using a multi-criteria fitness function that weighs residual energy, proximity to Sink device, and distribution density, and (3) an optimized path formation strategy for both intra-cluster and inter-cluster data transmission, leveraging an enhanced A*-based routing algorithm to minimize communication overhead and improve energy efficiency. Performance evaluations across three diverse scenarios demonstrate that EE-AIRP–averaged over multiple independent runs–achieves substantial energy-efficiency gains, approximately 40% relative to LEACH-C and 28%, 12%, and 6% compared with H-KDTREE, PECR, and KMSC, respectively. Moreover, the protocol extends network lifetime by promoting a more balanced distribution of energy consumption across sensor nodes than these baseline protocols. These findings–reported with dispersion measures–corroborate the robustness and reproducibility of EE-AIRP under both dense and sparse deployments. These improvements make EE-AIRP particularly suitable for IoT applications such as environmental monitoring, healthcare, and smart buildings, where network longevity is critical. The EE-AIRP code and corresponding simulation results can be found at: <span><span>https://doi.org/10.5281/zenodo.17149005</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"273 ","pages":"Article 111770"},"PeriodicalIF":4.6,"publicationDate":"2025-10-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145364295","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}