AI-Based Intrusion Detection Systems for IoMT Devices
Open Access DepositedWith the increased spread of Internet of Medical Things (IoMT) devices, healthcare infrastructure faces the risk of cyberattacks that threaten the privacy of patients, and the continued operation of healthcare. This study evaluates AI-based intrusion detection systems (IDS) for an IoMT environment using the WUSTL-EHMS-2020 dataset, and the comparative performance of feature selection methods. The effectiveness of these feature selection methods, Mutual Information and Shapley values, were investigated. In all models, it was found that Shapley values were able to produce better classification results. Four machine learning models were tested including Random Forest, XGBoost, Support Vector Machine, and an Autoencoder (for unsupervised anomaly detection). The overall accuracy, recall, and F1-scores were strongest for XGBoost, which supports the assumption of better performance. The autoencoder performed poorly when compared to Random Forest and Support Vector Machine, as the other two performed competently. Shapley value-based feature selection produced stronger classification performance, which supports that they can effectively be used as an interpretable and model-agnostic technique for IoMT cybersecurity. As a result, hybrid detection methods may be required to be used in the future. These findings contribute to existing research in machine learning for cybersecurity in healthcare and show the benefits of using Shapley value feature selection, and XGBoost, to detect harmful activity in IoMT systems. The results help medical device manufacturers, hospital IT groups, and support decisions when evaluating regulatory compliance in this field. There is an evidence-based framework for choosing interpretable, effective IDS models suited to healthcare’s special circumstances.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.