An intrusion detection system based on stacked ensemble learning for IoT network

As Internet of Things (IoT) technology advances rapidly, the need for security measures in this domain intensifies, mainly due to the inherent computational resource constraints and production cost limitations of IoT devices. These factors make them vulnerable to the botnet attacks. We propose an Intrusion Detection System (IDS) based on an effective stacked-based ensemble learning approach. Considering the computational resources constraints, we employ feature selection and hyperparameter optimization techniques to lessen the computational overhead and enhance detection performance. Our proposed IDS is tested on the Python platform and proven effective performance on NB-aIoT and UNSW-NB15 datasets. The experimental results show that our proposed IDS achieves a high average accuracy rate of 99.68% and has the potential to improve the security of IoT devices, ultimately benefiting users that rely on these devices.