Effectiveness of Machine Learning Based Intrusion Detection Systems

Abstract

Security is the most significant issue in concerns of protecting information or data breaches. Furthermore, attackers present a new variety of cyber-attacks in the market, which prevent users from managing their network or computer system. For that reason, the growth of cybersecurity research studies, such as intrusion detection and prevention systems have great significance. The intrusion detection system (IDS) is an effective approach against malicious attacks. In this work, a range of experiments has been carried out on seven machine learning algorithms by using the CICIDS2017 intrusion detection dataset. It ensued to compute several performance metrics to examine the selected algorithms. The experimental results demonstrated that the K-Nearest Neighbors (KNN) classifier outperformed in terms of precision, recall, accuracy, and F1-score as compared to other machine learning classifiers. Nevertheless, All of the used machine learning classifiers except KNN trained their models in a reasonable time.