Intelligent Cyber-Attack Detection in IoT Networks Using IDAOA-Based Wrapper Feature Selection

Abstract

ABSTRACT: In the realm of cybersecurity, the increasing sophistication of cyber-attacks demands the creation of sophisticated intrusion detection systems (IDS) designed to accurately detect and counteract threats in real-time. This study presents an innovative framework that integrates the Improved Dynamic Arithmetic Optimization Algorithm (IDAOA) with a Bagging technique to enhance the performance of intelligent cyber intrusion detection systems. The IDAOA serves as a wrapper-based feature selection method, optimizing the identification of the most impactful features while balancing local exploration and global exploitation. The Bagging technique further strengthens the classification phase by combining predictions from multiple classifiers, effectively addressing issues of class imbalance and improving overall system robustness. Evaluation of the proposed system using the NSL-KDD dataset demonstrates its superior performance, achieving an accuracy of 99.45%, significantly outperforming state-of-the-art approaches. These findings underscore the potential of intelligent optimization and ensemble learning techniques in advancing cybersecurity for IoT networks.