Network intrusion detection system using advanced feature selection and optimized deep neural networks for attack classification

Abstract

Recent advancements in network technologies have posed significant challenges to the performance of many existing Network Intrusion Detection Systems (NIDS). Most of these systems struggle to effectively integrate advanced feature selection methods, data imbalance handling techniques, and optimized deep neural network architectures, resulting in suboptimal classification performance. This research proposes novel Network Intrusion Detection System that integrates advanced feature selection, method (wrapper based techniques) data balancing method  and an optimized deep neural network (DNN) architecture using  bayesia optimization techniques  to enhance detection accuracy and overall system robustness. The proposed NIDS was evaluated using two widely recognized benchmark datasets namely NSL-KDD and   CIC-IDS2017—and compared against four intrusion detection models developed using state-of-the-art feature selection techniques  as well as other existing benchmark models. Experimental results reveal that the proposed model significantly outperforms the compared systems across key evaluation metrics by achieving accuracy improvement of greater than 0.05 value,along  side near perfect specifically 0.9674 on NSL-KDD and 0.9994 precision, recall, F1-score, demonstrating its superior capability in detecting and classifying network intrusions under diverse conditions.