Privacy-IoMT dataset

Abstract

Privacy-IoMT A Synthetic Benchmark for Privacy Violation Detection in the Internet of Medical Things Privacy-IoMT is a synthetic benchmark designed to support the development and evaluation of machine learning and anomaly-detection methods for privacy violation detection in Internet of Medical Things (IoMT) environments. Unlike conventional IoMT security datasets that primarily focus on cyberattacks and network intrusions, Privacy-IoMT models privacy-oriented access behaviors and violations across representative healthcare workflows. The benchmark is generated through a controlled simulation framework that produces normal and privacy-violating access events while preserving explicit relationships between healthcare context, access behavior, privacy constraints, and privacy-violation labels. Dataset Classes Privacy-IoMT contains the following primary classes: Normal – legitimate access behavior satisfying the defined privacy constraints. Unauthorized Access – access that violates authorization constraints. Abnormal Access – access behavior that violates expected access-frequency constraints. Insider Misuse – legitimate or privileged users accessing resources without an appropriate medical necessity. Data Leakage – access or communication involving an unauthorized or inappropriate data destination. The primary generative mapping between adversary/violation types and labels is intentional but not strictly exclusive. Some generated privacy-violation events may simultaneously satisfy multiple privacy constraints. In such cases, the final label is determined according to the priority-based classification rule defined in the accompanying methodology.