COMPARATIVE STUDY OF MACHINE LEARNING ALGORITHMS FOR INTRUSION DETECTION SYSTEMS

Abstract

Network environments must be safeguarded from a large number of cyber threats that are a real threat to most of the users. The development of ML techniques has brought about some noticeable improvements in the intrusion detection system (IDS) such as the ability to make better real-time analysis, adjust to newer concepts and provide a more accurate detection. This research involves the application of machine learning algorithms that use a comparative analysis to evaluate the performance of various IDS models. The research will discuss a range of machine learning techniques that are supervised and hybrid methods. We judge the models against time, precision, recall, etc., the main criterions in choosing the model. We find that while supervised machine learning models give high accuracy, using the Random Forests and SVM in a hybrid model improves the performance. Thus, the product is a hybrid model that combines the strengths of both approaches. For example, Random Forest can provide a strong feature representation but SVM can refine the decision boundary thus, lead to a more accurate and reliable classification model. Usually, this technique performs better than individual models or any single algorithm on its own.