Byzantine Attacker Identification in Collaborative Spectrum Sensing: A Robust Defense Framework

Abstract

The problem of Byzantine attack in collaborative spectrum sensing (CSS) is considered in this paper. To defend against Byzantine attack, a robust defense framework is proposed to efficiently identify the Byzantine attackers. Specifically, we first propose a robust defense framework, where a reference is built based on the extended sensing, and the transmit results and sensors are continuously evaluated via the reference and identified at intervals. In the framework, except of data falsification, multiple practical factors are considered, including the variation characteristic of sensors’ attributes, reporting channel imperfection, and inference errors based on the transmit results. Further, we derive the closed-form expressions of the reference and the identification performance and make optimization of the identification threshold in two cases: with and without the prior knowledge of attack behaviors, where the probability of correctly detecting Byzantine attackers is maximized under the constraint of the probability of falsely identifying honest sensors as attackers. In particular, when the prior knowledge is unavailable, maximized likelihood estimation is made based on the reference to achieve the optimization. Furthermore, we present in-depth simulations to demonstrate the high robustness of the proposed defence framework to multiple practical factors under a homogeneous scenario and a heterogeneous scenario.