Enhanced Intrusion Detection System Using Hybrid Harris Hawks Inspired Feature Selection For Efficient Network Intrusion Detection

Anomaly-based Intrusion Detection Systems (IDS) playa crucial role in identifying network breaches by detecting deviations from baseline patterns. However, existing methods often struggle to accurately differentiate between normal and intrusive activities, leading to false alarms and increased vulnerability. To address this limitation, this paper proposes a novel approach. Firstly, data preprocessing and normalization are conducted, followed by feature extraction using Stacked Convolutional Denoising Autoencoder (SCDAE). Subsequently, a Hybrid-Harris Hawks optimization (HHO) algorithm is used for feature selection. Next, a CNN-BILSTM-Attention architecture is developed for classification, with softmax and fully connected layers. The proposed approach achieves performance metrics on the NSL-KDD dataset, including an accuracy of 99.67%, precision of 99.74%, Fl-score of 99.65%, and Detection Rate of 99.69%. These results demonstrate the efficacy of the proposed method in overcoming the limitations of existing techniques such as Deep Neural Network (DNN), Decision Tree (DT), Bilateral Long Short-Term Memory (BiLSTM), and Random Forest (RF).