Computer Networks最新文献

{"title":"Generating P4 data planes using LLMs","authors":"Mihai-Valentin Dumitru,&nbsp;Vlad-Andrei Bădoiu,&nbsp;Alexandru M. Gherghescu,&nbsp;Costin Raiciu","doi":"10.1016/j.comnet.2025.111709","DOIUrl":"10.1016/j.comnet.2025.111709","url":null,"abstract":"<div><div>Over the past few years, Large Language Models (LLMs) have become the source of impressive results in code generation. However, most research focuses on widely adopted general-purpose programming languages, with little attention given to niche domain-specific languages (DSLs). This raises the question: do DSLs, such as P4, a data plane programming language, have a place in the LLM world?</div><div>The potential impact of generating DSL code could be tremendous. Automatically generating data plane code promises flexible networks that can quickly adapt to specific conditions at the lowest level. P4 is structurally simpler than general-purpose languages, but also offers a much smaller corpus of existing programs, thus setting up interesting challenges for deep-learning based code generation.</div><div>In this paper, we show that crafting a highly specialized P4 dataset with domain knowledge is sufficient to bootstrap P4 code generation through fine-tuning existing LLMs, even when they have not encountered P4 code during pre-training. We further document the process of creating a relevant benchmark to assess the proficiency of fine-tuned models in generating P4 code. Our evaluation shows that our fine-tuned models outperform much larger models in both syntactic correctness and semantic alignment.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"272 ","pages":"Article 111709"},"PeriodicalIF":4.6,"publicationDate":"2025-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145109816","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":"Joint design of sub-channel assignment and power control in D2D aided cellular system: a novel GNN and DRL based approach","authors":"Zhongyu Ma ,&nbsp;Ning Zhang ,&nbsp;Heng Zhang ,&nbsp;Yan Zhang ,&nbsp;Zhaobin Li ,&nbsp;Qun Guo","doi":"10.1016/j.comnet.2025.111708","DOIUrl":"10.1016/j.comnet.2025.111708","url":null,"abstract":"<div><div>Device-to-Device (D2D) communication integrated with cellular networks is viewed as a promising network technology for enhancement of power efficiency and spectral utilization in the proximity-based wireless applications scenarios. However, co-channel interference caused by simultaneous sharing of wireless resources between the concurrent links including D2D links and cellular links poses a significant challenge for this system. To this end, a novel graph neural network (GNN) and deep reinforcement learning (DRL) combined resource allocation framework is proposed in this paper. Firstly, the joint design of sub-channel assignment and power control in the D2D overlapped cellular system is investigated as intractable nonlinear programming, where the long-term sum-of-rate (LSR) of cellular links and the transmission success rate (TSR) of D2D links are simultaneously maximized under the constraints such as concurrent interference and traffic demands, etc. Secondly, a GNN and DRL combined resource allocation framework (GD-CRAF) is proposed, where the GNN based graph sampling and aggregation (GraphSAGE) is designed to efficiently exploit the interference features from the incomplete global interference graph constructed with local interference, and the double deep Q-network (DDQN) based sub-channel assignment and power control is proposed under the DRL framework. Finally, the superiority of the proposed GD-CRAF framework is verified in diversified scenarios, where the convergence and effectiveness of the GD-CRAF are demonstrated. It is shown from the experimental results that the LSR and TSR of the GD-CRAF are superior to that of other references such as DQN based scheme, Q-Learning based scheme and random allocation based scheme.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"272 ","pages":"Article 111708"},"PeriodicalIF":4.6,"publicationDate":"2025-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145106437","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":"Efficient key encapsulation mechanisms from noncommutative NTRU","authors":"Ali Raya ,&nbsp;Vikas Kumar ,&nbsp;Sugata Gangopadhyay ,&nbsp;Aditi Kar Gangopadhyay","doi":"10.1016/j.comnet.2025.111704","DOIUrl":"10.1016/j.comnet.2025.111704","url":null,"abstract":"<div><div>Key Encapsulation Mechanisms (KEMs) are cryptographic set of algorithms used to establish a shared secret between two parties over an insecure channel. In the context of post-quantum cryptography, KEMs are typically constructed from hard mathematical problems believed to resist quantum attacks. Among these, lattice-based schemes–particularly those based on the NTRU problem–have been widely studied due to their efficiency and strong security foundations. Traditional NTRU constructions operate over commutative polynomial rings, offering a good balance between speed and compactness. However, recent efforts have proposed noncommutative variants of NTRU to enhance resistance against algebraic attacks. While these variants improve security properties, they generally fall short in terms of performance when compared to the original NTRU. This work introduces the first noncommutative NTRU construction that matches the performance of classical NTRU over the ring of integers. In addition, we propose a second design based on the ring of Eisenstein integers, further enhancing computational efficiency. We provide full KEM implementations of both constructions and benchmark them against existing commutative and noncommutative NTRU-based schemes. Our results demonstrate that the twisted dihedral group ring-based construction achieves encapsulation and decapsulation speeds on par with NTRU-HPS, while improving key generation speed by a factor of 2.5. The Eisenstein integer-based scheme shows an improvement of 1.6<span><math><mo>×</mo></math></span> in key generation and 1.3<span><math><mo>×</mo></math></span> in encapsulation and decapsulation. These findings confirm that noncommutative algebra can be leveraged effectively to achieve competitive performance in practical post-quantum KEM designs.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"272 ","pages":"Article 111704"},"PeriodicalIF":4.6,"publicationDate":"2025-09-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145109812","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":"Data-driven resource allocation for ensuring remote data collection timeliness in integrated ground-air-space networks","authors":"Jinsong Gui ,&nbsp;Hanjian Liu","doi":"10.1016/j.comnet.2025.111715","DOIUrl":"10.1016/j.comnet.2025.111715","url":null,"abstract":"<div><div>Ensuring remote data collection timeliness without terrestrial network infrastructure support is a huge challenge. The exploration of addressing this challenge with the aid of opportunistic unmanned aerial vehicles (UAVs) and satellites has received extensive attention. In this paper, we address a data-driven resource allocation problem, which aims to ensure data collection timeliness, minimize communication resource waste, and maximize data collection amount under the UAVs’ opportunistic access mode and satellites’ random access mode. However, due to UAVs’ dynamic behaviors, time-varying data collection missions, real-time matching demand between ground nodes and UAVs, and free competition of UAV-satellite access resources, it will be difficult to achieve the above goal if it is considered as a global optimization problem. Thus, we construct three problems in turn that collectively describe the requirements of above goal, and then reformulate the first two problems as the Markov decision process models and take deep reinforcement learning tools to get the corresponding solutions, respectively. Next, the solution to the third problem is approximated by alternately invoking the algorithms of the first two problems. Finally, our simulation results are compared with those of other benchmark schemes from different perspectives, and the effectiveness and superiority of the presented solutions are verified.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"272 ","pages":"Article 111715"},"PeriodicalIF":4.6,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145106440","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":"Dynamic eMBB scheduling strategy for GBR and NGBR in Non Standalone 5G NR: A deep reinforcement learning approach","authors":"H. Eddine Benmadani ,&nbsp;M. Amine Ouamri ,&nbsp;M. Azni ,&nbsp;T. Essa Alharbi","doi":"10.1016/j.comnet.2025.111692","DOIUrl":"10.1016/j.comnet.2025.111692","url":null,"abstract":"<div><div>With the emergence of 5G networks and network slicing concept, efficient resource management is crucial to meet varied Quality of Service (QoS) requirements. Intra-slice scheduling plays a central role in optimizing the network performance while catering to the requirements of different traffic flows within a slice. In this paper, we propose a Deep Reinforcement Learning (DRL)-based scheduling scheme for eMBB applications. This appraoch aims to maximize system throughput, increase GBR throughput, minimize packet loss by minimizing Head of Line (HoL) delay, and ensure NGBR flow fairness. To evaluate our approach, we test and contrast the two DRL methods, i.e., Deep Q-Network (DQN) and Proximal Policy Optimization (PPO). Using both approaches to fine-tune our hybrid scheduling metric, we demonstrate the adaptability and reliability of our approach with different learning frameworks. We contrast our DRL-based scheduler performance to the Proportional Fair (PF) scheduler and two QoS-aware schedulers, QoS and EXP-PF. Simulation shows that our scheme significantly improves the system throughput and maintains the GBR and NGBR traffic performance in balance. Moreover, the comparison of DQN and PPO provides novel insights into wireless scheduling efficacy with a foundation for future adaptive scheduling solutions in 5G and beyond.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"272 ","pages":"Article 111692"},"PeriodicalIF":4.6,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145061108","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":"Trust the Source: A latency-based machine learning approach to accurate IP geolocation in internet","authors":"Miguel A. Ortega-Velázquez ,&nbsp;Alejandro S. Martínez-Sala ,&nbsp;Pilar Manzanares-López ,&nbsp;Maria-Dolores Cano ,&nbsp;Antonio J. Jara","doi":"10.1016/j.comnet.2025.111721","DOIUrl":"10.1016/j.comnet.2025.111721","url":null,"abstract":"<div><div>IP geolocation is the process of determining the geographic location of an Internet-connected device based on its IP address. Ensuring the authenticity of data sources has become critical for robust cybersecurity and plays a vital role in safeguarding systems by enabling applications such as fraud prevention, cybercrime investigations, and location-based access controls. There are two main approaches to IP geolocation: passive methods, which rely on public or historical data but may be outdated or inaccurate; and active methods, which use real-time latency measurements or routing path topology to infer location. Inspired by wireless location systems and the fingerprinting technique, this work proposes an active IP geolocation system that leverages Machine Learning to estimate IP locations using Round-Trip Time (RTT) latency measurements taken from a distributed network of probing nodes, referred to as Monitors. A central Coordinator collects RTT data from Monitors pinging known landmarks to build RTT fingerprints. These are used to train ML models that infer the location of unknown target nodes. The testbed system, consisting of a Coordinator server and six Monitors distributed across Europe, operated over a 65-day measurement campaign. More than 2 million RTT samples were collected from approximately 1700 Landmarks (used to train/test the ML models) and 1200 targets (used to evaluate the system). The K-Nearest Neighbours (KNN) and Multi-Layer Perceptron (MLP) algorithms are considered and compared with the reference Constraint-Based Geolocation (CBG) approach. The evaluation finds that the proposed system is capable of geolocating a point with a mean error of 317.6 km, a 38 % reduction compared to the CBG baseline. On the other hand, the average delay to complete the geolocation process is less than 5 s. These results demonstrate a scalable and cost-effective solution for medium-grained accuracy and bounded-delay IP geolocation in cybersecurity contexts.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"272 ","pages":"Article 111721"},"PeriodicalIF":4.6,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145106007","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 comprehensive survey of DDoS attack defense systems for different SDN architectures","authors":"Mitali Sinha","doi":"10.1016/j.comnet.2025.111711","DOIUrl":"10.1016/j.comnet.2025.111711","url":null,"abstract":"<div><div>Software-Defined Networking (SDN) is gaining popularity as the new generation networking platform across diverse domains such as 5G, IoT, and cloud computing. Its widespread acceptance is due to the innovative principle of decoupling the network’s control logic from its data-forwarding hardware. This decoupling allows network administrators to dynamically configure and manage network resources through software, providing unparalleled flexibility and agility. SDN has two types of architectures: pure SDN and hybrid SDN, each designed to meet specific requirements like pure SDN is often used in environments where there is a need for dynamic network management, such as data centers and cloud computing environments, hybrid SDN is commonly implemented in existing network infrastructures where organizations want to gradually adopt SDN without completely overhauling their network architecture. This study aims to present a comprehensive survey of Distributed Denial of Service (DDoS) attack defense systems for different types of SDN architectures. Specifically, this research (a) classifies DDoS defense systems based on the SDN architectures and conducts a comparative analysis of existing studies for each architecture, (b) develops a set of guidelines to enhance current DDoS defense solutions, and (c) identifies several future research directions for designing DDoS defense mechanisms against emerging DDoS attack types in the context of SDN. This work is distinct from previous studies as DDoS defense solutions are analyzed based on the specific architectures of SDN, an aspect not addressed in prior surveys.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"272 ","pages":"Article 111711"},"PeriodicalIF":4.6,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145106438","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":"TitNet: A time-series model based on multi-period nesting for encrypted traffic classification","authors":"Congcong Wang ,&nbsp;Xin Li ,&nbsp;Zhaoqiang Cui ,&nbsp;Lina Xu ,&nbsp;Jiangang Hou ,&nbsp;Jie Sun ,&nbsp;Hongji Xu ,&nbsp;Zhi Liu","doi":"10.1016/j.comnet.2025.111702","DOIUrl":"10.1016/j.comnet.2025.111702","url":null,"abstract":"<div><div>Encrypted traffic classification is essential for network management tasks such as quality-of-service controls, identifying malicious traffic, and enhancing cybersecurity. However, the scarcity of plaintext information and the significant reduction of payload characteristics in encrypted traffic present challenges to effective classification. To tackle these issues, we propose a novel time series model called TitNet, which models network traffic at the session level as a multivariate time series and effectively integrates periodic and spatial features inherent in time series data. Our TitNet contains a dynamic frequency selection strategy(DFSS) that facilitates the conversion of time series data into two-dimensional tensor representations, which is pivotal for accurately discerning the intricate patterns embedded in encrypted traffic. This approach enables TitNet to iteratively transform time series into 2D tensors, effectively exploiting the multi-period nesting characteristics of the data to improve classification performance. Experimental results on the ISCXTor2016 dataset (43 Tor/NonTor categories) robustly indicate that our TitNet excels in the detection, classification, and identification of applications within encrypted traffic, achieving 96.21 % accuracy while handling extreme class imbalance. Nonetheless, TitNet introduces additional computational overhead and relies on fixed session truncation, which may limit scalability and long-range modeling. Future work will explore lightweight variants and improved sequence aggregation strategies to address these challenges.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"272 ","pages":"Article 111702"},"PeriodicalIF":4.6,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145158526","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":"Autoencoder-based decentralized federated learning for efficient communication","authors":"Abdul Wahab Mamond,&nbsp;Majid Kundroo,&nbsp;Taehong Kim","doi":"10.1016/j.comnet.2025.111676","DOIUrl":"10.1016/j.comnet.2025.111676","url":null,"abstract":"<div><div>Decentralized federated learning (DFL) has emerged as a solution for traditional federated learning’s limitations, such as network bottlenecks and single-point failure, by enabling direct communication between nodes and eliminating the reliance on a central server. However, DFL still encounters challenges like increased communication costs as the number of participating nodes increases, amplifying the need for efficient compression techniques. Moreover, the increasing complexity of models, including vision, language, and generative models (e.g., GPT), further underscores this necessity due to their large parameter sizes. To address the communication cost-related issues in DFL, this study introduces Autoencoder-based Decentralized Federated Learning (AEDFL), which leverages autoencoders to compress model updates before transmission, allowing them to be reconstructed at the receiving end with high fidelity and minimal loss of accuracy. We conduct comprehensive experiments using two models, SqueezeNet and DenseNet, on three benchmark datasets: CIFAR-10 (under both IID and non-IID settings), FashionMNIST, and CIFAR-100. The results demonstrate that AEDFL achieves up to 122x compression with negligible accuracy degradation, showcasing its effectiveness in balancing communication efficiency and model performance across varying model sizes and dataset complexities.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"272 ","pages":"Article 111676"},"PeriodicalIF":4.6,"publicationDate":"2025-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145106008","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":"Adaptive memory replay for network intrusion detection: Tackling data drift and catastrophic forgetting","authors":"Nasreen Fathima A H ,&nbsp;Ansam Khraisat ,&nbsp;Syed Ibrahim S P ,&nbsp;Gang Li","doi":"10.1016/j.comnet.2025.111712","DOIUrl":"10.1016/j.comnet.2025.111712","url":null,"abstract":"<div><div>Network intrusion detection aims to identify anomalous activities in network traffic, while continual learning (CL) methods strive to preserve past knowledge and adapt to evolving threats. Memory replay-based CL approaches have been widely used and proven effective at mitigating catastrophic forgetting. However, previous research has primarily focused on addressing class imbalance and has largely relied on augmented and random memory replay strategies, which introduce significant computational overhead and limit practicality in real-time applications. To overcome these challenges, we propose Task-Aware Memory Replay (TAMR), a novel framework that prioritizes past experiences based on their relevance to the current task. By dynamically adjusting the importance of replayed samples, TAMR balances the integration of new attack patterns with the retention of critical historical knowledge, ensuring resilience against evolving threats and variations in normal traffic. Unlike traditional methods that employ random selection or augmented replays, TAMR selectively replays high-impact experiences, thereby optimizing memory usage and improving adaptability. Our experiments demonstrate that TAMR achieves real-time adaptability across five distinct NIDS datasets, ultimately delivering superior performance and computational efficiency in detecting even unknown attacks in dynamic network environments. In general, we highlight the potential of memory-based replay strategies for continual learning in detecting unknown attacks using a task-aware approach.</div></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":"272 ","pages":"Article 111712"},"PeriodicalIF":4.6,"publicationDate":"2025-09-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145109815","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}