Performance Analysis of Machine Learning Techniques in Intrusion Detection

Abstract

This paper presents the performance analysis of machine learning techniques in intrusion detection. We analyze time to build (and to retrain) the models used by Intrusion Detection System. Machine Learning is a branch of computer science that allows computer to learn by themselves without programming sequence. These techniques can be applied to detect new threat that has never seen before. Due to the large volumes of security audit data as well as complex and dynamic properties of intrusion behaviors, optimizing the accuracy of IDS becomes an important open problem that is receiving attentions from the research community. However, the performance (time and space required) is usually ignored. Our study allows administrators work to make better decisions about how to select the proper hardware for intrusion detection in various environments. We proposed the models for estimating the time to build each model and the vector equation of the cut-off point is provided for determining the minimum number of CPU required for building Decision tree model and support vector machine model.