Multivariate Mutual Information-based Feature Selection for Cyber Intrusion Detection

Abstract

Cyber security is one of the most serious threats for security of the large-scale network such as smart grids. An effective and fast cyber intrusion detection is paramount for reliable performance of the system. Proper and efficient feature selection is one of the most important issues in cyber intrusion detection. In this paper, a novel multivariate mutual information based feature selection (MVMIFS) is proposed to select the most relevant and important features for intrusion detection. Least square support vector machine (LSSVM) is used to classify the traffic data with high accuracy. The proposed method is validated in three well-known datasets; KDD Cup 99, NSL-KDD, and Kyoto 2006 $+$ . The experimental results show that the proposed method outperforms existing approaches in detection rate, accuracy and false positive rates.