Machine Learning-Based Intrusion Detection: A Comparative Study

Abstract

Recently, the internet use is expanded leads for many type of attacks on the network.  As a result, a robust and effective network intrusion detection system is needed to strengthen its defense and performance. The main purpose of the intrusion detection system remains to monitor and analyze the system process for potential malicious acts committed by hackers. As a result, researchers have conducted several reviews on such topics, but most of these studies were not comprehensive. In this paper, the authors create a machine learning-based intrusion detection system and use a robust and close set of attribute selection methods with classifiers using a group review, and analyzing of attribute-choosing methods common with functions. The study extracts the important attributes from continuous variables by applying attribute-choosing methods to generate an important variable set and an intrusion detection system. KDD data were double-checked to obtain outcomes from this process. The performance results clearly showed the mathematical k-nearest neighbor’s (K-NN) algorithm outperforms the other classifiers. It was also noted that the use of attribute choosing techniques based on the percentage of information gain is preferable compared to other features.