Network intrusion detection system using supervised learning paradigm

Abstract

Internet has positively changed social, political and economic structures and in many ways obviating geographical boundaries. The enormous contributions of Internet to business transactions coupled with its ease of use has resulted in increased number of internet users and consequently, intruders. It is crucial to safeguard computer resources with the aid of Intrusion Detection Systems (IDS) in addition to Intrusion Prevention Systems. In recent times, enormous network traffic generated in terabytes within couples of seconds are difficult to analyze with the traditional rule-based approach; hence, researchers have to subject data mining techniques to intrusion detection with emphasis on intrusion detection accuracy; relevant feature selection leads to faster and enhanced accurate detection rate. Therefore, this paper presents a light weight IDS based on information gain and Multi-layer perceptron Neural Network. Gain ratio was used in selecting relevant features for attack and normal traffic prior classification using Neural Network. Empirical results from the UNSW-NB15 intrusion detection dataset on thirty selected attributes is a highly ranked decision, thus, the light weight IDS is suitable for real time intrusion detection.