Information Content Analysis and Clustering for Signal Anomaly Detection

An information theoretic approach to detecting unusual events in radio signals is presented. Anomalies are detected through a measure of the events' information content. Clustering is utilised to reduce false-positives while allowing a lower discrimination threshold to be used for improved anomaly detection. Experiments with a real wireless local area network test signal shows that it is possible to achieve 100% detection rates while maintaining very low false discovery rate of 1%.