Case Study II: HMM-Based Byzantine Attack Detection

Abstract

Most of the existing defense schemes against the Byzantine attack discussed in Chap. 3 either assume that the underlying spectrum states at different timeslots are independent or they only focus on the measurements collected in a single timeslot. Nonetheless, for many practical scenarios, the activities of the PUs and the induced spectrum states often follow a Markov process, and hence the spectrum sensing behaviors of the SUs may be better characterized by the HMM. Under this modeling, a novel HMM-based Byzantine attack detection technique can be developed to enforce the robustness of collaborative spectrum sensing. To illustrate this, the HMM-based spectrum sensing behavioral model is presented first, and based on which, the sought-after multi-HMM inference algorithm is introduced. Then, the overall HMM-based Byzantine attack detection scheme is demonstrated along with some numerical results to corroborate its effectiveness.