A Hybrid Classifier Approach for Network Intrusion Detection

Abstract

In this paper, we have proposed a new hybrid classifier approach for detecting the incoming signal as normal or attack in a network. Before applying the hybrid approach we have solved the class imbalance problem using sampling techniques so as to improve the accuracy of detection. After balancing the data set it combines three approaches namely Support Vector Machine (SVM), decision tree (J48) and Naive Bayes (NB). NSL-KDD data set, the improved version of KDDCUP'99 data set was used to measure the performance of our hybrid approach based on certain features. SVM was used to extract the required features from the data set. It is observed that applying hybrid approach after balancing the data set gives promising results and improves the accuracy of detection.