Enhanced Intrusion Detection Based Hybrid Meta-heuristic Feature Selection

As technology advances and becomes more complex, the frequency and complexity of cyber-attacks also increase. Unscrupulous hackers and cybercriminals are developing innovative methods to penetrate computer systems and unlawfully steal crucial data. To address these threats, organizations must implement effective Intrusion Detection Systems (IDSs) to detect and respond to assaults swiftly. IDSs have recently attracted significant attention due to their effectiveness and precision in detecting abnormal patterns in network traffic using machine learning techniques. This work aims to create a model that can identify intrusions by employing various machine learning algorithms on the chosen features obtained from the modelling process. Evolutionary algorithms and local search techniques are merged in hybrid mode to build new models to select the most suitable features and prepare them for the proposed investigation. The SMOTE method is used for addressing the issue of imbalanced datasets. The utilization of imbalance techniques is highly effective. It may effectively rectify the uneven distribution of classes, hence enhancing the accuracy of a machine learning model in identifying the underrepresented class. A study is conducted to investigate various machine learning techniques and assess the effectiveness of the proposed methodology. The evaluation of research on the widely recognized Intrusion Detection benchmark CSE-CIC-IDS2018 and KDD CUP 99 datasets demonstrates significant promise.