Network Intrusion System Detection Using Machine and Deep Learning Models: A Comparative Study

Abstract

As Internet technology advances, the exponential growth of network security risks is becoming increasingly apparent. Safeguarding the network poses a significant challenge in the realm of network security. Numerous security measures have been put in place to detect and pinpoint any malicious activities occurring within the network. Among these mechanisms, the Intrusion Detection System (IDS) stands out as a widely utilized mechanism to reduce the impact of these risks. In this paper, a comparative study of different machine learning classifiers namely decision tree, random forest, SVM, KNN and different deep learning models including CNN, GRU and LSTM is performed to select the best approach for intrusion detection. Our approach consists of three main steps: pre-processing, feature extraction and selection, and classification and validation. The experiments are conducted using two large datasets namely KDDCUP99 and Hogzilla datasets in both binary and multi-class classification fashion. The evaluation results show the superiority of deep learning methods over machine learning classifiers on both datasets.