Scholarlink

Cognitive radio network (CRN) has progressed over the conventional radio to effectively employ the unused frequency spectrum. CRN is vulnerable to several threats during spectrum sensing access. Different methods are utilized to minimize anomaly attacks as malicious users degrade network performance. However, the dynamic and heterogeneous nature of spectrum usage has significant difficulties in CRN for anomaly detection. In this research, the self-attention recurrent reinforcement learning (SARRL) is proposed to detect anomalies in CRN. Initially, the CSV file dataset is considered which is available in Kaggle to determine the proposed SARRL performance. The standard scalar is applied in the preprocessing phase which increases model performance. The SARRL is used to detect the anomalies effectively by self-attention mechanism in CRL. Accuracy, F1-score, recall, and precision are utilized in this research as performance measures. The proposed SARRL achieves a better probability detection of 0.99 compared to existing techniques like gated recurrent unit-support vector machine (GRU-SVM), respectively.

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

Self-Attention Recurrent Reinforcement Learning Based Anomaly Detection for Dynamic Spectrum Access in Cognitive Radio Networks

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Sachinkumar,
  • A. Suresh,
  • Rajeshwari Patil,
  • Alampally Sreedevi,
  • Yerrolla Chanti

摘要

Cognitive radio network (CRN) has progressed over the conventional radio to effectively employ the unused frequency spectrum. CRN is vulnerable to several threats during spectrum sensing access. Different methods are utilized to minimize anomaly attacks as malicious users degrade network performance. However, the dynamic and heterogeneous nature of spectrum usage has significant difficulties in CRN for anomaly detection. In this research, the self-attention recurrent reinforcement learning (SARRL) is proposed to detect anomalies in CRN. Initially, the CSV file dataset is considered which is available in Kaggle to determine the proposed SARRL performance. The standard scalar is applied in the preprocessing phase which increases model performance. The SARRL is used to detect the anomalies effectively by self-attention mechanism in CRL. Accuracy, F1-score, recall, and precision are utilized in this research as performance measures. The proposed SARRL achieves a better probability detection of 0.99 compared to existing techniques like gated recurrent unit-support vector machine (GRU-SVM), respectively.