Optimized detection of cyber-attacks on IoT networks via hybrid deep learning models

Abstract

The rapid expansion of Internet of Things (IoT) devices has significantly increased the potential for cyber-attacks, making effective detection methods crucial for securing IoT networks. This paper presents a novel approach for detecting cyber-attacks in IoT environments by combining Self-Organizing Maps (SOMs), Deep Belief Networks (DBNs), and Autoencoders. These techniques are employed to create a system capable of identifying both known and previously unseen attack patterns. A comprehensive experimental framework is established to evaluate the methodology using both simulated and real-world traffic data. The models are fine-tuned using Particle Swarm Optimization (PSO) to achieve optimal performance. The system’s effectiveness is assessed using standard cybersecurity metrics, with results showing an accuracy of up to 99.99% and Matthews Correlation Coefficient (MCC) values exceeding 99.50%. Experiments conducted on three well-established datasets NSL-KDD, UNSW-NB15, and CICIoT2023 demonstrate the model’s strong performance in detecting various attack types. These findings suggest that the proposed approach can significantly enhance the security of IoT systems by accurately identifying emerging threats and adapting to evolving attack strategies.