Refined Intrusion Detection Model for Internet of Things Networks

The expansion of the Internet of Things (IoT) has permeated various sectors, facilitating the integration of IP connectivity into many devices. These devices, also known as Internet of Things devices, are vulnerable to a diverse array of security threats. As a result, there has been a significant increase in the occurrence of IoT malware cases in the past few years. Protecting IoT devices from such assaults has become an urgent and crucial topic. However, conventional peripheral security measures are insufficient to deal with the lightweight security requirements inherent to the IoT landscape. Considering this, we propose a Refined Intrusion Detection Model for IoT networks (RIDM-IoT) that demonstrates similar efficiency in exposing malicious activities compared to the existing computationally expensive methods. The primary strength of the proposed model is that it's capable of efficiently detecting attacks in IoT networks while requiring less computational resources. RIDM-IoT accomplishes this by employing a unique hybrid feature selection strategy that combines filter-based and wrapper-based feature selection methods. As a wrapper-based feature selection approach, we employ the well-known genetic algorithm, Particle Swarm Optimisation (PSO). The suggested feature selection approach decreases the size of the feature space by 93%. Furthermore, we employ simple machine learning algorithms to construct intrusion detection models instead of intricate deep learning models. We conducted tests on RIDM-IoT with unseen attack types to verify its general behaviour. The results demonstrate that the suggested model effectively detects attacks in IoT networks, using a smaller collection of features when compared to state-of-the-art complex models.