Performance evaluation of the fuzzy ARTMAP for network intrusion detection

Abstract

Recently, considerable research work have been conducted towards finding fast and accurate pattern classifiers for training Intrusion Detection Systems (IDSs). This paper proposes using the so called Fuzzy ARTMAT classifier to detect intrusions in computer network. Our investigation shows, through simulations, how efficient such a classifier can be when used as the learning mechanism of a typical IDS. The promising evaluation results in terms of both detection accuracy and training duration indicate that the Fuzzy ARTMAP is indeed viable for this sort of application.