Labelling clusters in an intrusion detection system using a combination of clustering evaluation techniques

Abstract

A new clusters labelling strategy, which combines the computation of the Davies-Bouldin index of the clustering and the centroid diameters of the clusters is proposed for application in anomaly based intrusion detection systems (IDS). The aim of such a strategy is to detect compact clusters containing very similar vectors and these are highly likely to be attack vectors. Experimental results comparing the effectiveness of a multiple classifier IDS with such a labelling strategy and that of the classical cardinality labelling based IDS show that the proposed strategy behaves much better in a heavily attacked environment where massive attacks are present. The parameters of the labelling algorithm can be varied in order to adapt to the conditions in the monitored network.