A comparative analysis of gradient boosting, random forest and deep neural networks in intrusion detection system

Abstract

The growing threat of advanced security attacks targeting enterprise information systems raises the need for novel security solutions that promptly identify and respond to these issues. These security strategies must automate threat detection and response in enterprise settings, enabling organizations to address emerging threats, ongoing attacks, and imminent risks adequately. Traditional security strategies that rely on rule-based approaches for intrusion detection systems are inefficient in achieving these objectives due to their limited capabilities in identifying new threats. As a result, machine learning strategies have been proposed to address these needs, offering an intelligent detection environment for novel threats. Classification algorithms such as random forest, gradient boosting and deep learning techniques like deep neural networks have been proposed in various studies. This paper examines the performance of these models, providing a comparative review of their detection capabilities based on precision, recall, accuracy, specificity, and sensitivity. The models are tested using a Python environment due to the extensive machine learning capabilities. These tests show that random forest is the ideal model for network-based intrusion detection systems