A Comparison of Machine Learning Classifiers for Network Intrusion Detection System

Abstract

The primary objective of this paper is to assess and detect intrusions, which is one of the most complicated tasks due to the increasing diversity of attacks. As advanced breaches grow increasingly, it will become more difficult to detect them in various industries, such as industry and national security. Traditional intrusion detection methods are no longer capable of detecting malicious behaviour that follows unusual patterns. CSE-CICIDS2018 dataset is a popular dataset used for testing intrusion detection systems (IDS). This research was intended to develop predictive models for network-based intrusion detection. That is the latest intrusion detection dataset, which is huge data, open source, and covers a broad spectrum of attack patterns. This research uses two machine-learning-based algorithms, the Random Forest and Decision Tree algorithms, to focus on training and testing accuracy of the dataset. This paper finds out that the Random Forest provides the highest 99% accuracy as compared to the Decision Tree.