A New Similarity Measure for the Anomaly Intrusion Detection

Abstract

This paper introduces a new similarity measure that can be applied for the anomaly intrusion detection by using weighted complete bipartite graphs. The first set of nodes represents users, while the second set depicts the characteristics defining his profile. The weight on each edge is computed from the frequency of appearances of characteristics for a given user. We demonstrate the validity of our measure by fulfilling the set of rules defined for any similarity measure.