Intrusion Detection Based On Clustering Algorithm

Abstract- The traditional Intrusion detection systems have been used long time ago, namely Anomaly-Based detection and Signature-based detection but have many drawbacks that limit their performance. Consequently the main goal of this paper is to use data mining techniques including classification using clustering methods to overpass these defects. This classification will be done by using k-means algorithm. Therefore we have improved k-means to overcome its limits specially the cluster’s number initialization. The experiment results of the work done on KDD’99 dataset shows the performance of the improved k-means in detecting new attacks with more than 90 % for Dos and R2L also more than 60 % for probe and U2R