A lightweight deep learning model for intrusion detection in Internet of Medical Things (IoMT) devices using weight pruning technique

Abstract

The Internet of Medical Things (IoMT) has transformed modern healthcare by enabling seamless data collection and monitoring through connected medical devices. However, the growing interconnectivity of these devices exposes them to serious cyber security threats, such as unauthorized access, data tampering, and denial-of-service attacks. These risks are heightened by the limited computing and memory resources of IoMT devices, which make traditional intrusion detection systems unsuitable. This study proposes a lightweight deep neural network model optimized for resource-constrained IoMT environments. The model integrates Principal Component Analysis (PCA) for dimensionality reduction and weight pruning techniques to minimize model complexity while maintaining high detection performance. Experimental results demonstrate a significant reduction in model size to 84 KB with 98% detection accuracy, outperforming state-of-the-art methods. This research provides an efficient and deployable security solution that strengthens the resilience of IoMT devices without overwhelming their limited computational capacity