Abstract: Journal of Material Sciences & Applied Engineering

Abstract:
Intrusion Detection Systems (IDSs) are essential for securing highly confidential data and protecting network architecture from cyber-attacks. Despite their high accuracy, traditional IDS methods often face significant challenges, such as model bias due to data imbalances and irrelevant features. This study proposes the state-of-the-art machine learning (ML) based IDS that addresses these challenges. By minimizing misclassification errors and correcting model bias, this proposed IDS significantly enhances predictive accuracy and generalizability, thereby offering a promising solution to the current limitations of IDS technologies. This study used the Decision Tree, Xtreme Gradient Boosting, and Adaboost model to classify an attack. The experimental results demonstrate the robustness of a XGB model for the classification of an attack.