Publication

Abstract : In recent years, with the popularity of fitness bands and smartwatches, there has been a surge in the collection of personal physiological data. Tracking activities and fitness levels has become an intrinsic part of many people's lives. While the deep insight it provides into an individual's health has multiple benefits, this lifestyle choice comes with the risk of endangering the user's privacy. Data from these devices has previously been proven capable to identify a person, much like a fingerprint. Most existing privacy attacks rely on fine-grained multi-sensor data. Through our observations we can demonstrate that pedometer data, collected per default by iPhones, even when aggregated to a 15-minute resolution, can be used to identify a user with good precision if sufficient training data is used to create the fingerprints. We also demonstrate that the participant's data has such a strong individual signal that they can be clustered with high accuracy if they are divided into days. © 2022 Elsevier Ltd