Performance Analysis of Deep Learning Models for IoMT Intrusion Detection

Abstract: The rapid expansion of the Internet of Medical Things (IoMT) has improved healthcare delivery but has also increased exposure to sophisticated cyber-attacks. To detect these attacks, this study investigates a deep learning–based intrusion detection analysis tailored to IoMT-enabled healthcare infrastructure. Six promising deep learning architectures are systematically evaluated under both binary (benign vs. malicious) and multiclass classification settings along with model efficiency by analysing computational complexity and inference latency. The analysis incorporates healthcare-specific preprocessing, spatial and temporal feature learning, and class imbalance mitigation. Performance assessed through accuracy, precision, recall, F1-score, and complexity analysis. Experimental results demonstrate that temporal and hybrid models achieve near perfect binary accuracy and better multiclass performance, outperforming existing IoT and IoMT studies. The findings highlight the effectiveness and deployment potential of deep learning-based IDS for securing real world smart healthcare environments. Keywords: Internet of Medical Things, Healthcare cybersecurity, Intrusion detection systems, Hybrid and Deep Learning Models, Computational efficiency, CIC-IoMT-2024.