Advances in IoT Intrusion Detection: Deploying Hybrid Deep Learning and Metaheuristic Algorithms for Optimal Feature Selection
Ahmed B. Abdulkareem · Ingénierie des systèmes d information · 2025
The proliferation of the Internet of Things (IoT) has tremendously increased the attack vectors for cyber threats, which necessitates advanced intrusion detection systems.In this paper, we propose a new method for detecting intrusions to the IoT based on a new hybrid deep learning model and metaheuristic algorithm for optimal feature selection.Our methodology takes advantage of the synergy between the CNN, BiGRU, and BiLSTM grids of models and integrates them into one architecture that enables them to leverage the spatial and temporal attributes of data to improve anomaly detection.While the model is refined using the Genetic, Harris Hawk, Dragonfly, Grey Wolf, and Particle Swarm Techniques, PSO demonstrates the best results, with a 98.11% accuracy level.This research uses comprehensive ToN-IoT datasets for analysis, which includes a wide variety of normal and adversarial traffic patterns affecting the IoT.The results show that our new hybrid model not only possesses a high level of accuracy, but it also exhibits considerable potential for real-world deployment.We further suggest potential areas for developing the model, such as scalability, real-time readiness, and its integration in environment computing.Our study advances cybersecurity by developing a cost-effective solution that can provide optimal protection to the IoT from multiple intrusion cases.