A Hybrid Grey Wolf Optimizer and Deep Transfer Learning-Based Intrusion Detection System for IoT

Abstract

The rapid proliferation of Internet of Things (IoT) devices has driven significant advancements in connectivity and automation, but it has also introduced substantial cybersecurity risks, including intrusions and cyberattacks. To address these challenges, this study proposes a hybrid intrusion detection framework that combines advanced optimization techniques with ensemble deep learning to enhance detection accuracy and computational efficiency. The proposed framework integrates Grey Wolf Optimizer (GWO) with Tabu Search for feature selection, effectively eliminating irrelevant and redundant features. These refined features are processed using an ensemble deep learning model comprising stacked Long Short-Term Memory (LSTM) networks and Gated Recurrent Units (GRU), which are adept at capturing complex temporal patterns in network traffic data. The framework was rigorously evaluated on benchmark datasets, including NSL-KDD, UNSW-NB15, and Edge-IIoT. The model achieved results, with accuracies of 99.55, 98.57 and 95.08%, respectively. Additionally, the system demonstrated a superior sensitivity of 99.21% and specificity of 99.45% on the NSL-KDD dataset while achieving high precision rates across all datasets. Comparative analyses showed that the framework consistently outperformed models such as GWO-GRU, GWO-LSTM, and GWOTB-LSTM in terms of accuracy, detection rate, and false alarm reduction. This robust and adaptable framework addresses the unique challenges of intrusion detection in IoT networks, supporting secure deployment in industrial and consumer IoT systems.