Scholarlink

<p>Abstract</p><p>Cognitive radio wireless sensor networks (CR-WSNs) are particularly susceptible to routing vulnerabilities arising from dynamic-spectrum availability and sophisticated adversarial attacks, emphasizing the need for secure and efficient routing mechanisms. Current solutions address routing optimization, spectrum management, and security as independent tasks, leading to suboptimal performance and susceptibility to Byzantine jamming, spectrum sensing data falsification (SSDF), and primary user emulation attacks (PUEA). This article introduces DRL-SecRoute (deep reinforcement learning-based secure routing), a new unified secure routing framework that synergistically integrates deep reinforcement learning with adaptive spectrum sensing to address the multi-dimensional optimization problem of secure routing in dynamic CR-WSN environments. The key contributions are fourfold: (1) a twin-delayed deep deterministic policy gradient (TD3) algorithm enhanced with prioritized experience replay (PER), specifically designed for continuous state-action spaces in CR-WSNs, achieving 40% faster convergence than existing discrete-action methods; (2) a hybrid adaptive spectrum sensing mechanism combining Bayesian inference with Kalman filtering that reduces sensing overhead by 37.4% while maintaining high prediction accuracy; (3) a lightweight multi-layered anomaly detection system integrating statistical divergence analysis with unsupervised learning (Isolation Forest) to detect diverse attacks with 91.7% accuracy and less than 6% false positive rates; and (4) a multi-objective optimization framework that jointly optimizes routing latency, energy consumption, spectrum efficiency, and security risk through the unified DRL approach. Extensive simulation experiments across varying network densities (50–200 nodes), traffic loads, spectrum activity levels, and six attack scenarios—including an adaptive reinforcement learning-based adversary—demonstrate that DRL-SecRoute achieves a packet delivery ratio of up to 97.8% under collaborative attack conditions and sustains 84.2% PDR even against an adaptive reinforcement learning-based adversary, improves energy efficiency by 32.7% over existing protocols, maintains a spectrum access collision rate below 3.5% across all primary user activity levels, and extends network lifetime by 41.3%. These results confirm that DRL-SecRoute offers a reliable and scalable foundation for next-generation CR-WSN deployments.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

DRL-SecRoute: a synergetic deep reinforcement learning paradigm for mitigating byzantine faults and SSDF attacks through heuristic spectrum cognizance in next-generation cognitive radio networks

  • 期刊论文

  • 发表日期：

    2026年5月6日

查看全文
  • Manish Kumar Dixit,
  • Arun Kumar Singh,
  • Saurabh Dixit

摘要

Abstract

Cognitive radio wireless sensor networks (CR-WSNs) are particularly susceptible to routing vulnerabilities arising from dynamic-spectrum availability and sophisticated adversarial attacks, emphasizing the need for secure and efficient routing mechanisms. Current solutions address routing optimization, spectrum management, and security as independent tasks, leading to suboptimal performance and susceptibility to Byzantine jamming, spectrum sensing data falsification (SSDF), and primary user emulation attacks (PUEA). This article introduces DRL-SecRoute (deep reinforcement learning-based secure routing), a new unified secure routing framework that synergistically integrates deep reinforcement learning with adaptive spectrum sensing to address the multi-dimensional optimization problem of secure routing in dynamic CR-WSN environments. The key contributions are fourfold: (1) a twin-delayed deep deterministic policy gradient (TD3) algorithm enhanced with prioritized experience replay (PER), specifically designed for continuous state-action spaces in CR-WSNs, achieving 40% faster convergence than existing discrete-action methods; (2) a hybrid adaptive spectrum sensing mechanism combining Bayesian inference with Kalman filtering that reduces sensing overhead by 37.4% while maintaining high prediction accuracy; (3) a lightweight multi-layered anomaly detection system integrating statistical divergence analysis with unsupervised learning (Isolation Forest) to detect diverse attacks with 91.7% accuracy and less than 6% false positive rates; and (4) a multi-objective optimization framework that jointly optimizes routing latency, energy consumption, spectrum efficiency, and security risk through the unified DRL approach. Extensive simulation experiments across varying network densities (50–200 nodes), traffic loads, spectrum activity levels, and six attack scenarios—including an adaptive reinforcement learning-based adversary—demonstrate that DRL-SecRoute achieves a packet delivery ratio of up to 97.8% under collaborative attack conditions and sustains 84.2% PDR even against an adaptive reinforcement learning-based adversary, improves energy efficiency by 32.7% over existing protocols, maintains a spectrum access collision rate below 3.5% across all primary user activity levels, and extends network lifetime by 41.3%. These results confirm that DRL-SecRoute offers a reliable and scalable foundation for next-generation CR-WSN deployments.