Optimized Intrusion Detection Using Harris Hawks Feature Selection and Deep Learning: A High-Performance Approach

The rapid development of complex network devices and technology has increased the need for effective intrusion detection systems (IDS) to battle complex cyberattacks. In this research, a novel hybrid method is proposed by combining Harris Hawks Optimization (HHO) for feature selection and a two-layer deep learning framework to enhance intrusion detection. We use the benchmark dataset NSL-KDD dataset, then the most efficient features are chosen by the HHO algorithm, and the deep learning model, with emphasis on efficiency and simplicity, achieves high accuracy in classification. Our approach attains a very good accuracy of 99.34%, outperforming many recent methods, and demonstrating competitive results compared to complex architectures. With the use of HHO for optimal feature selection and the addition of dropout layers for preventing overfitting, the proposed method ensures scalability, resilience, and applicability in the real world of intrusion detection. This work showcases the promise of combining intelligent optimization and deep learning in addressing prominent security challenges and establishes a baseline for the pursuit of ensemble approaches in the future.