Intrusion System Detection System using Decision Tree Compared to Linear Regression

Abstract

IDS plays a very much important role in network security by detecting and counteracting the malicious activities happening. This paper analyzes which one of two specified machine learning algorithms-working on network intrusion detection, shows better performance-Decision Tree (DT) or Linear Regression (LR). We used a good dataset as well as applied both algorithms on detections for numerous intrusions kinds. The accuracy achieved by the Decision Tree is much higher, at 94.7%, along with precision at 92.5% and recall at 91.8%. Where Linear Regression got to an accuracy of 82.3%, a precision of 79.6%, and a recall of 78.9%, evidently that algorithm has not been successful in identifying intricate intrusion patterns correctly. To classify non-linear and high-dimension data, it is proven that Decision Tree performs better. Our results show that while DT is generally more viable for complex robust IDS solutions, LR can, on the other hand, be applicable to less complex scenarios with lower classification complexities. The findings from this study are useful for the purpose of appropriateness in terms of algorithm selection based on the network environment and the complexity of the attack to further improve the performance of the intrusion detection.