Defense Against Byzantine Attacks: Anomaly Detection Using One-Class SVM in Cooperative Spectrum Sensing

Cognitive radio has emerged as a viable solution to address the challenge of spectrum scarcity in modern wireless communication systems. Spectrum sensing, a crucial component of cognitive radio, often encounters difficulties associated with wireless channel impairments like path loss, fading, and shadowing when performed using single-user detection methods. Cooperative spectrum sensing (CSS) has been introduced to mitigate these adverse channel effects. However, CSS is susceptible to manipulation by malicious users (MUs), posing a significant obstacle to the success of cognitive radio. Therefore, detecting MUs within the cognitive radio network (CRN) is essential to enhance the detection performance of CSS. This study proposes a one-class support vector machine (SVM) based anomaly detection algorithm for the MUs detection. The proposed algorithm blindly detects MUs in the CRN without requiring prior knowledge of attacker behavior. We also demonstrate the performance of our proposed algorithm in detecting attackers across different attack scenarios and compare it with the different state-of-the-art algorithms. Furthermore, leveraging the MU detection algorithm, we propose a weighted sum-based CSS algorithm capable of mitigating the impact of attackers on CSS performance, ultimately improving the reliability and security of spectrum-sharing in real-world applications.