Byzantine Defense in Collaborative Spectrum Sensing via Bayesian Learning

Collaborative spectrum sensing (CSS) enables secondary users in cognitive radio networks to collaboratively explore spectrum holes as well as protecting the primary user from being interfered. Unfortunately, the emergence of spectrum sensing data falsification (SSDF) attack, also known as the Byzantine attack, brings significant threat to the reliability of the CSS. Majority of the existing studies on Byzantine defense can be divided into two categories: one is directly to make the judgment based on the current spectrum sensing data, while the other uses the historical spectrum sensing data to update sensors' reputation. The first category of studies does not take the historical spectrum sensing data into account, while most of the second category of studies are heuristic in nature. In this paper, we invoke Bayesian learning to design Byzantine defense schemes. First, we develop a Bayesian offline learning algorithm by considering one practical challenge that the ground-truth spectrum state is unavailable for training. Then, we develop a Bayesian online learning algorithm by considering the case that the sensors' attribute may be time-varying. In addition, we present simulations to show the performance of the proposed defence algorithms.