A Fast Malicious User Detection Scheme Based on POMDP for Cooperative Spectrum Sensing in Cognitive Radio networks

Cooperative spectrum sensing (CSS) can improve spectrum sensing accuracy, but it can be injured due to potential attacks from malicious cognitive radio user who reports false sensing results to the fusion center (FC). Many researchers focus on reducing the effect of malicious users on the accuracy of spectrum sensing. A promising method to detect malicious users is to determine their abnormal spectrum sensing behavior. In this paper, we provide a novel malicious users detection scheme for cognitive radio (CR) based on the truth rate of each CU, which is defined as the correlation level between the Markov property of the CU’s reported sensing information and the states of the PU signal. The truth rate may distinguish an honest user from a malicious user by giving an honest CU a high trust rate and giving a malicious user a low one. In the malicious user detection process, a partially observable Markov decision process (POMDP) is applied to consider the effect of the current action (that action is to classify a CU as an honest or a malicious user) on the reward in future time slot (that reward is achieved by classifying a CU as an honest or a malicious user). By taking advantage of POMDP, the proposed scheme may detect the presence of malicious users in a shorter required time.