A Multi-Class Intrusion Detection System for the Internet of Medical Things Based on Hybrid Deep Learning

Abstract

The IoMT improves healthcare through smart medical devices, enabling real-time monitoring and data transmission. However, increased connectivity exposes IoMT systems to cyber threats, jeopardizing patient data confidentiality, system integrity, and availability. Traditional IDS struggle to detect sophisticated attacks, thus requiring advanced solutions. This study presents a hybrid deep learning model that integrates LSTM and DNN to improve intrusion detection in IoMT networks. The CICIoMT2024 dataset, comprising network traffic of 40 IoMT devices under 18 types of cyberattacks, was used for training and evaluation. Data preprocessing included label encoding, normalization. The LSTM component captures sequential traffic patterns, while the DNN extracts advanced features for classification. Batch normalization, dropout layers, and early stopping were implemented to improve model performance. Experimental results show that the proposed model outperforms the conventional intrusion detection system, achieving 99.6% accuracy in binary classification, 99.4% in 6-class classification, and 98.4% in 19-class classification. Compared with stand-alone models, the hybrid approach demonstrates superior accuracy and robustness. This research underscores the effectiveness of LSTM-DNN in securing IoMT networks. Future work will focus on real-time deployment, optimization of computational efficiency, and expansion of the dataset to improve cyber threat detection in medical settings.