Optimizing Intrusion Detection with a Dual-Archive Approach Using DAPS-XG

Intrusion Detection Systems are essential for protecting industrial networks, critical infrastructure, and corporate environments against deliberate cyber-attacks. Nevertheless, a notable obstacle in contemporary intrusion detection systems is the need to handle extensive, multi-dimensional data effectively while ensuring good detection precision and efficient processing. This paper presents DAPS-XG, a sophisticated intrusion detection model that combines Multi-Objective Particle Swarm Optimization for selecting features with XGBoost for categorization. A prominent characteristic of DAPS-XG is its dual-archive Multi-Objective Particle Swarm Optimization system, which integrates two separate archives: a convergence archive, aimed at optimizing classification accuracy, and a diversity archive, intended to investigate a broad spectrum of feature subsets. This dual-archive system effectively optimizes the trade-off between improving detection performance and minimizing computing complexity. The model underwent evaluation using the NSL-KDD dataset, which is a known benchmark in the field of intrusion detection research. The DAPS-XG algorithm effectively decreased the dimensionality of the dataset from 41 to 18 essential features, resulting in a remarkable F1 score of 99.55%. The model is especially suitable for real-time intrusion detection in industrial settings, where high-speed and precise analysis of substantial amounts of network traffic data is crucial. Through the substantial reduction of processing requirements, while maintaining detection accuracy, DAPS-XG provides a very efficient solution for enhancing IDS performance in intricate, data-intensive settings, rendering it extremely applicable to modern cybersecurity applications.