An Improved Medoid Clustering Algorithm For Intrusion Detection Using Web Usage Mining Technique

Abstract

Intrusion detection is one of the applications of web usage mining. In this application, we find the intrusive or worthless data through mining techniques, determine the user behaviour, i.e. new user or existing user, label data according to the users' requirements and detect networks' known and unknown attacks. There are various models of detection of intrusion. Misuse and anomaly detection are types of intrusion. In anomaly detection, the intrusion is unknown and known in misuse. There are various techniques that we discuss in this paper.

We proposed a novel algorithm for intrusion detection using mining techniques based on the medoids and means clustering algorithm. We also compared the proposed algorithm with existing algorithms with high detection and low false alarm rates to detect known and unknown attacks.