Wrapper-based Feature Selection for Enhanced Intrusion Detection Using Random Forest Classification

In network environments, detecting and mitigating cyber-attacks necessitates the creation of an effective Intrusion Detection System. This research describes an IDS framework that combines Random Forest (RF) classification and a wrapper-based feature selection strategy to increase performance. The wrapper-based strategy iteratively picks the most relevant features by lowering dimensionality and removing redundant or irrelevant data, increasing the RF classifier's efficiency. The framework is assessed against three well-known benchmark datasets: NSL-KDD, CICIDS-2019, and Bot-IoT. Experimental results show that when used in conjunction with wrapper-based feature selection, the RF classifier improves detection accuracy, precision, and recall significantly. This solution handles the complexity and variety of modern network traffic, resulting in a more accurate and efficient IDS than older systems.