Goodness-of-Fit-Based Malicious User Detection in Cooperative Spectrum Sensing

Cooperative spectrum sensing improves sensing accuracy in primary user detection, but can be threatened by malicious users. Malicious users may try to falsify the sensing result to indicate that the primary user exists even when there is no primary user in order to monopolize the spectrum usage, thereby depriving other users of their spectrum opportunities. To address this, we propose a malicious user detection scheme where the malicious users are identified and cut off from the cooperative sensing process. The proposed scheme exploits the Anderson-Darling (AD) goodness-of-fit technique which tests whether the empirical distribution of the sensing data from each secondary user fits the expected distribution for a malicious user. In addition, we derive false alarm and detection probabilities for when malicious users are cut off by the malicious user detection scheme. Simulation results show that the proposed goodness-of-fit-based malicious user detection significantly improves sensing performance in comparison with conventional outlier detection-based schemes.