"Digital Watermarking of Medical Sensor Data for Data Leakage Detection - a Proof-of-Concept Prototype"
, in Masterarbeit am Institut für Wirtschaftsinformatik - Data & Knowledge Engineering, Betreuung: o. Univ.-Prof. Dr. Michael Schrefl, unter Anleitung von Mag. Dr. Bernd Neumayr, 9-2020
Digital Watermarking of Medical Sensor Data for Data Leakage Detection - a Proof-of-Concept Prototype
Sprache des Titels:
Medical devices produce medical sensor data which are time indexed sequences of medical measurements. By using a data platform, patients can permanently store their medical sensor data and share it with authorized data users such as doctors or researchers. But authorized data users may leak accessed data to unauthorized parties. Digital watermarking can be used to imperceptibly watermark requested data with a watermark associated to the requesting data user. If leaked data is found anywhere, the leaking data user can be identified by extracting the embedded watermark. In this thesis, we introduce a new watermarking approach for medical sensor data to detect data leakage. Due to medical sensor data being permanently stored in the data platform, watermark embedding and detection are informed, i.e. they take advantage of original data. In addition, watermark embedding is configurable by usability constraints such that watermarked data remains useful for diagnostic purposes. A major challenge for watermark detection are malicious attacks which aim to distort or remove embedded watermarks. In order to counteract malicious attacks, watermark detection is based on similarity searches. In this thesis, we design, implement and evaluate a proof-of-concept prototype which is based on the introduced watermarking approach. Among other things, the evaluation shows that the prototype can be considered secure against common variants of distortion, collusion, subset selection and deletion attacks. Furthermore, the evaluation shows that the performance of the prototype can be considered sufficient for practical use of providing continuous blood glucose measurements for data users such as doctors. The prototype may also be applied to different kinds of sensor data and provides high adaptability as well as extensibility.