Efficient Hierarchical ML-Based IoT Intrusion Detection System Leveraging PSO and Sequential Forward Feature Selection

The Internet of Things (IoT) is the main factor behind the success of the technological era in the fourth industrial revolution. However, the development of IoT is facing various security-related challenges, which is the most dangerous problem. To address this, an efficient hierarchical Intrusion Detection System (IDS) based on ensemble Machine Learning algorithms is proposed, with two layers conducting binary and multiclass classification. Particle Swarm Optimization and sequential forward feature selection are also used to extract two optimal feature sets to enhance the outcome and reduce the complexity of the proposed IDS solution. As a result, two sets of features, including 5 and 4 features extracted from the original IoT-23 dataset, are selected for each layer, achieving 99.94% Accuracy (Acc) and 99.76% F1-Score in binary classification and 99.25% Acc in multiclass classification, respectively. The proposed method accurately detects intrusions and identifies specific attacks, helping to solve security issues in IoT environments.