A hybrid feature reduced approach for intrusion detection system

Abstract

Increase in the number of Internet users every day leads to the global concern of information security. Intrusion detection plays an important role in protecting data from any intruders that occur in our networks or system. It detects the problem by observing and analyzing the events occurring in the system. Most of the intrusion detection techniques focus on the feature reduction or selection in order to improve the generalization ability. Feature reduction aims at finding the most compact and informative set of features by removing worthless and redundant data or information from the database. The proposed study uses conventional principal component analysis (PCA) with information gain (IG) and chi-square (CHI) for feature reduction. The reduced features are then used to perform classification using support vector classifier. The validation of the proposed approach has been done on the KDD-99 dataset and promising results have been obtained.