Selecting features for anomaly intrusion detection: A novel method using fuzzy C means and decision tree classification

Abstract

In this work, a new method for classification is proposed consisting of a combination of feature selection, normalization, fuzzy C means clustering algorithm and C4.5 decision tree algorithm. The aim of this method is to improve the performance of the classifier by using selected features. The fuzzy C means clustering method is used to partition the training instances into clusters. On each cluster, we build a decision tree using C4.5 algorithm. Experiments on the KDD CUP 99 data set shows that our proposed method in detecting intrusion achieves better performance while reducing the relevant features by more than 80%. © Springer International Publishing Switzerland 2013.