Fine Decision Tree Outperforms in Early Intrusion Detection: A Supervised Learning Comparison on NSL-KDD and NID

Abstract

The intrusion detection system plays a major role towards network security, to prevent network threats. An intrusion detection system is used to keep track of network or systems activity in a bid to detect any mischievous activity. The dataset containing intrusion attacks is used to detect abnormalities in the network with the help of the machine learning algorithm. Machine learning has three major subcategories, which include supervised, unsupervised, as well as reinforcement learning. The most common and the most important supervised learning classifiers are used in machine learning. The most scholar has been looking on the intrusion detection through various machine-learning methods. Nevertheless, it still brings out some weaknesses. In order to identify the most successful supervised machine learning algorithm that could be used in detecting intrusions. This paper used the two feature-based datasets, namely NSL_KDD and NID and the five supervised machine learning algorithms, such as Support Vector Machine, Naive Bayes, Logistics Regression, Decision Tree and Neural Networks. Those algorithms can be used in an intrusion detection system, however, the comparison of the results demonstrates their efficiency. It also recommends the best approach which should be adopted so as to prevent attacks at an early stage in this field. The fine Decision Tree algorithm had an accuracy of 99.4%, which is better in identifying intrusions at the beginning. Their performance is far much better when compared to other algorithms.