Intrusion Detection System Using Unsupervised Immune Network Clustering with Reduced Features

Intrusion Detection Systems (IDS) are developed to be the defenseagainst security threats. Current signature based IDS like firewalls andanti viruses, which rely on labeled training data, generally cannot detectnovel attacks. The purpose of this study is to enhance the detection rateby reducing the network traffic features and to investigate the feasibilityof bio-inspired Immune Network approach for clustering different kindsof attacks and some novel attacks. Rough Set method was applied toreduce the dimension of features in DARPA KDD Cup 1999 intrusiondetection dataset. Immune Network clustering was then applied usingaiNet algorithm to cluster the data. Empirical study revealed thatdetection rate was enhanced when most significant features were used torepresent input data. The finding also revealed that Immune Networkclustering method is robust in detecting novel attacks in the absence oflabels