Performance evaluation of data clustering techniques using KDD Cup-99 Intrusion detection data set

Abstract

Intrusion detection systems aim at detecting attacks against
computer systems and networks or, in general, against information systems.
A number of techniques are available for intrusion detection. Data mining is
the one of the efficient technique among them. Intrusion detection and
clustering have forever been hot topics in the area of machine learning. Data
clustering is a procedure of putting related data into groups. Clustering
procedure clusters the data into groups with the property of inter-group
similarity and intra-group dissimilarity. A clustering technique partitions a
data-set into several groups such that the likeness within a group is larger
than amongst groups. Clustering as an intrusion detection technique has long
before proved to be beneficial.

This paper evaluates four most representative off-line clustering
techniques: k-means clustering, fuzzy c-means clustering, Mountain
clustering, and Subtractive-clustering. These techniques are implemented and
tested against KDD cup-99 data set, which is used as a standard benchmark
data set for intrusion detection. Performance and accuracy of the four
techniques are presented and compared in this paper. Results shows Accuracy
of K-means is 91.02%, FCM is 91.89%, Mountain clustering is 75% and
Subtractive clustering is 78.27%. The experimental outcomes obtained by
applying these algorithms on KDD cup-99 data set demonstrate that k-means
and fuzzy c-means clustering algorithms perform well in terms of accuracy
and computation time.