INTRUSION DETECTION SYSTEM WITH SUPERVISED LEARNING AND FEATURE SELECTION

Abstract

Two machine learning techniques, namely SVM (Assist Vector Machine) and ANN (Artificial Neural Networks), are evaluated in this study for their efficiency. The presence or absence of regular or anomalous signatures in the request data will be determined using machine learning techniques. Internet intrusion detection systems (IDS) monitor request data and check if it contains typical or assault signatures; if it does, demand is reduced. This is necessary because nowadays all services are accessible online and malicious individuals can attack client or web server machines through this net. When new request signatures come, the IDS will be educated with all possible strikes signatures using AI techniques and then used to determine whether the new request comprises regular or assault trademarks. Here, we compare and contrast two AI algorithms, Support Vector Machines (SVM) and Artificial Neural Networks (ANN), and we find, experimentally, that ANN is more accurate than the current state-of-the-art SVM. Examining how well SVM and ANN work is the focus of this article. By utilising Relationship Based and Chi-Square Based function option formulas, the author has reduced the dataset dimension, eliminated irrelevant data, and loaded the model with important attributes. As a result of these features choice formulas, the dataset dimension will decrease and the forecast accuracy will increase.