Performance Analysis of an Anomaly Detection Method Based on Distributed Spatial Filtering

Abstract

This letter analyzes the performance of an anomaly detection method based on distributed spatial filtering, i.e., the spatial filtering of signals over networks in a distributed manner. Assuming that the signal value of each node is corrupted by noise, we evaluate the success probability of the anomaly detection. By focusing on the effects of the noise on the signal values and using appropriate approximations, we derive an approximate expression for the success probability as a function of the threshold to detect the anomaly. Moreover, we present a design method for the threshold to maximize the approximate success probability.