ABSTRACT:

In recent years, there has been an increase in the number of networked computers, making them open to various forms of cyber attack. Machine learning-based Intrusion Detection Systems (IDS) have become increasingly popular as a means of protecting networks from these types of risks. The effectiveness of present IDSs, particularly for less common attack types, is limited by issues such as stale and unbalanced datasets. We propose developing and evaluating an IDS based on machine learning techniques including K-Nearest Neighbour, Random Forest, Support Vector Machine, and Decision Tree. We use the recent, realistic, and imbalanced CSE-CIC-IDS2018 dataset. Experimental results show that our method considerably increases the detection rate for infrequently seen intrusions, making IDSs more effective and efficient against modern cyber threats.

Keywords: Intrusion detection.knn,svm, CSE-CIC-IDS2018 dataset,DT,RM