An Explainable Machine Learning Framework for Anomaly Detection in Internet of Medical Things (IoMT) Data
An Explainable Machine Learning Framework for Anomaly Detection in Internet of Medical Things (IoMT) Data
mahla baratpoor1 Hassan Abdullahi2
1) Department of Health Information Sciences, School of Management and Medical Information Sciences, Kerman University of Medical Sciences, Kerman, Iran.
2) Department of Health Information Sciences, School of Management and Medical Information Sciences, Islamic Azad University, South Tehran Branch, Tehran, Iran.
IOI Code
:
AAHD-ABGHZE
Publication :
National Conference on Information Technology, Nanotechnology, Artificial Intelligence and Technological Futures Studies(itnaf.ir)
Abstract :
The Internet of Medical Things (IoMT), as a core enabler of digital transformation in healthcare, has facilitated continuous patient monitoring and the real-time collection of biomedical data through interconnected devices. While this technology offers substantial benefits for improving diagnosis, treatment, and patient care, its dependence on networked infrastructures also exposes it to a wide range of cybersecurity threats. Consequently, anomaly detection in IoMT data has emerged as a critical strategy for identifying abnormal behaviors and mitigating both security-related and clinical risks.
This study presents a conceptual framework for anomaly detection in IoMT environments through the integration of deep learning and explainable artificial intelligence (XAI). In the proposed framework, simulated data derived from vital signal patterns and environmental sensor measurements were used to represent normal, noisy, and anomalous conditions in a controlled setting. The core detection mechanism is based on an autoencoder model, which learns compact latent representations of normal data patterns and detects anomalies using reconstruction error as the primary decision criterion. Furthermore, to enhance transparency and support user trust, the SHAP method was employed to explain the contribution of individual features to the generation of anomaly alerts.
The conceptual evaluation results indicate that the proposed framework outperforms several baseline methods in capturing nonlinear data patterns, maintaining robustness under noisy conditions, and reducing false-positive alerts. In addition, the incorporation of XAI significantly enhances the interpretability of the system, thereby improving its reliability and practical applicability in sensitive medical and cybersecurity contexts. Overall, the proposed approach provides a promising foundation for the development of transparent, robust, and trustworthy anomaly detection systems in next-generation IoMT ecosystems.
Keywords : internet of medical things (iomt) time series anomaly detection autoencoder explainable ai (xai) shap cybersecurity