Feature Selection Algorithm for Constructing Intrusion Detection System

Abstract

Abundance and unnecessary features in the large amount of data have caused a problem in data traffic classification which in turn slowdowns the classification process. Not only does this it also not allow the classifier for making extract decisions, which play a major role in big data. This system uses an algorithm based on mutual information which in turn selects the optimal features for classifications analytically, since it can handle linear and non linear features. Its efficiency can be evaluated in the network detection system. An Intrusion Detection System (IDS) named Least Square Support Vector Machine is fabricated using the feature selected by the algorithm. The performance of LSSVM-IDS can be obtained using three kinds of dataset namely KDD Cup 99, NSL-KDD and Kyoto 2006 dataset. The results show that algorithm contributes more critical features for the LSSVM-IDS to accomplish better exactness and lower computational cost.