A Nearest Neighbour-Based Analysis to Identify Patients from Continuous Glucose Monitor Data
Continuous glucose monitors (CGMs) are minimally invasive sensors that detect blood glucose levels (usually in patients with diabetes) at high frequency. The devices produce considerable volumes of sensor data when used for weeks and months. We consider the following research question: is it possible to uniquely identify a patient from a fragment of their CGM data? That is, supposing a patient’s medical records are stored in a database along with a large sample of their CGM data, could an attacker with a much smaller sample of data from a different time period match the two time series and positively identify the patient? If the answer is yes, then significant patient privacy concerns are raised since many health records are now stored online. Our investigations using existing public CGM datasets reveal that many subjects can be uniquely identified using a simple nearest neighbour-based analysis approach.
KeywordsContinuous glucose monitors Diabetes Nearest neighbour analysis Time series data Privacy Medical Internet of Things Data security
- 5.Hassanalieragh, M., et al.: Health monitoring and management using Internet-of-Things (IoT) sensing with cloud-based processing: opportunities and challenges. In: 2015 IEEE International Conference on Services Computing (SCC), pp. 285–292. IEEE (2015)Google Scholar
- 11.The Nightscout Foundation: About the Nightscout Data Commons on Open Humans (2014). http://www.nightscoutfoundation.org/data-commons/