Optimizing Feature Selection for Machine Learning-Based Intrusion Detection Systems Against Modern Cybersecurity Threats

Abstract

Technologies like AI, cloud computing, and big data have come a long way and changed a lot for the better. But this rise of cyberattacks to ensure an effective intrusion detection system (IDS). The challenges include lower accuracy from redundant features, less ability to detect new attacks from a single machine learning classifier, high rates of false alarms (FAR), excessive building time of models, etc.This paper introduces a hybrid feature selection approach with an ensemble classifier to select relevant features and give consistent classification of attacks. As the most recent open-access IDS dataset, the CICIDS-2017 dataset has significant promise as a prospective benchmark for IDS of the future since it incorporates contemporary system configurations and threat profiles. Yet in research, especially feature selection, it is still yet to be fully utilized. To overcome these issues, this study introduces a novel IDS framework deploying ensemble-based feature selection to derive a low-dimensional feature subset, and an ensemble-based IDS model for benchmarking on CICIDS-2017. The proposed scheme is a valuable contribution to the research community by integrating the most recent IDS dataset with ensemble methods for feature selection and detection, offering a strong solution for contemporary network security.Keywords: Cyber-Physical Systems (CPS), Intrusion Detection Systems (IDS), Ensemble-Based Methods, Feature Selection, Cybersecurity Threats