Machine learning and feature selection approach for anomaly based intrusion detection: A systematic novice approach

Abstract

Network Intrusion Detection System (NIDS) has become an imminent research area in network and information security due to the proliferation of the Internet and rapid increase in anomalous activities or intrusions. NIDS helps to detect anomalous activities or intrusions which compromise CIA (confidentiality, integrity, and availability), violate the security policies and mechanisms of a computer network. This paper presents a survey on anomaly based NIDS using machine learning technique employing feature selection approach. The prime contribution of this research is to present technical and empirical evaluation of each paper. The state-of-the-art NIDS is systematically analyzed and discussed according to machine learning and feature selection techniques used, number of selected features, efficiency in terms of various performance metrics and its result. This paper also provides an idea of selecting more appropriate solution and also the scope of improvement for each specific case.