Defending Against Byzantine Attack in Cooperative Spectrum Sensing: Defense Reference and Performance Analysis

Cooperative spectrum sensing (CSS) has been well recognized as an effective method to improve spectrum sensing accuracy and decrease sensing devices' complexity. However, spectrum sensing data falsification attack, also known as Byzantine attack, poses critical threats on the reliability of CSS. Due to lack of the ground-truth spectrum state, a reliable defense reference is vital to identify malicious behaviors and perform effective data fusion. However, the existing defense references have strong assumptions such as the attackers are in minority and/or a trusted node exists for data fusion. This observation motivates this paper to propose a novel defense reference, which jointly exploits the cognitive process of spectrum sensing and spectrum access in a closed-loop manner, to provide the defense scheme a solid basis without requiring any prior knowledge. Moreover, this paper analyzes the proposed reference's favorable reliability and high robustness over the state-of-the-art references, from two perspectives of spectrum sensing performance and the capability of identifying malicious sensors, respectively. Next, we design an optimal cooperative spectrum sensing scheme based on the proposed defense reference. Remarkably, from an information theoretic perspective, it is observed that based on the proposed reference, the information value of falsified reports is also exploited to further improve the global sensing performance. Furthermore, numerical simulations verify the proposed scheme's favorable performance, even in critical cases when malicious sensors are in majority.