H-C-LSTM: A Hybrid Deep Learning Model for Robust Intrusion Detection in IoMT Systems

The high rate of incorporation of the Internet of Medical Things (IoMT) into healthcare organizations has transformed patient monitoring, diagnosis, and treatment. There is, however, a growing risk of critical cybersecurity vulnerability associated with the growing interconnection of medical devices, which may lead to unauthorized access to sensitive data, service failures, and even manipulation of life-critical devices. Conventional intrusion detection systems (IDS) cannot cope with peculiarities of IoMT networks, i.e., the heterogeneity of devices, the lack of resources, and changing attack patterns. To overcome these issues, we introduce the Hybrid Convolutional Long Short-Term Memory (H-C-LSTM) model which is a hybrid of the Convolutional Neural Networks (CNN) to extract spatial features and the Long Short-Term Memory (LSTM) networks to model the sequences of features of time. Combining these two deep learning methods, our model helps to capture both spatial and temporal dependencies, which contributes to the process of detecting multi-stage and complex cyberattacks in the context of the IoMT. When assessing the H-C-LSTM model with real-world IoMT data, e.g., CICIoMT2024 and Bot-IoT, we have shown that the model has increased performance with an accuracy of 99.8, precision of 98.9; recall of 98.3; and an F1-score of 98.6. According to our findings, H-C-LSTM model can be used as a scalable, efficient, and robust model to secure the IoMT systems against the changing cyber threats when compared to traditional and recent deep learning-based models.