ENHANCING INTRUSION DETECTION SYSTEMS' ROBUSTNESS FROM ADVERSARIAL ATTACKS USING METAHEURISTIC ALGORITHMS

Abstract

Intrusion Detection Systems (IDS) play a crucial role in safeguarding computer networks from malicious activities. The increasing prevalence of malicious attacks on IDS bases systems has posed a significant threat to their robustness and effectiveness. It has become essential to build strong defense procedure to confirm the transparency and dependability of these systems. This research introduces an efficient methodology that incorporates ensemble learning technique like GBM, genetic operators, DNN-Deep Neural Network and Particle Swarm Optimization (PSO) with defense system to enhance the robustness of IDS against non-friendly system attacks. In order to maximize the selection of the most appropriate features from the given dataset, the suggested methodology starts with a feature engineering stage that makes use of PSO and GBM. In order to create a robust and discriminative feature subset, genetic operators are then used to further recheck the process of selection using different features.