Detecting Abnormal Power Emission for Orderly Spectrum Usage

This correspondence investigates the detection of a common kind of spectrum misuse behaviors, i.e., abnormal power emission, which may be induced by various factors, e.g., device fault, selfish motivation, or malicious attack. This problem is challenging due to the lack of prior information on specific abnormal behaviors and the uncertainty on real-time normal spectrum states. In this paper, we first formulate the detection of abnormal power emission as a composite binary hypothesis test. Then, a generalized likelihood ratio detector is derived to achieve the Neyman–Pearson criterion. Further, to simplify the detector and facilitate the performance analysis, a two-step detector called Normal before Abnormal detector is developed, and the corresponding detection performance and decision region are theoretically analyzed. Finally, simulation results show that the proposed detectors can well distinguish states with abnormal powers from normal spectrum states.