An Efficient Feature Redundancy Removal Approach towards Intrusion Detection in Ad Hoc Network

Abstract

Intrusion detection is a critical component of secure information systems. Data Intrusion Detection Processing System often contains a lot of redundancy and noise features, bringing the system a large amount of computing resources, a long training time, a poor real-time, and a bad detection rate. For high dimensional data, feature selection can find the information-rich feature subset, thus enhance the classification accuracy and efficiency. Based on a improved feature selection algorithm, this paper proposes a lightweight intrusion detection model with computational efficiency and high detection accuracy. The algorithm is based on information gain and SVM. Its principle is to group all data features according to information gain, and then to choose the feature subset with the best classification accuracy according to SVM algorithm(the classification accuracy of SVM is defined as intrusion Detection accuracy). The experimental results demonstrated that our approach can find features subsets with higher classification accuracy compared with feature selection algorithm based on information gain and GA.