Identification issues associated with the use of wearable accelerometers in lifelogging

Abstract

Personal lifelogging builds upon the pervasive and continuous acquisition of sensor measurements and signals in time, and this may expose the subject, and eventually bystanders, to privacy violations. While the issue is easy to understand for image and video data, the risks associated to the use of wearable accelerometers is less clear and may be underestimated. This work addresses the problem of understanding if acceleration measurements collected from the wrist, by subjects performing different types of Activities of Daily Living (ADLs), may release personal details, for example about their gender or age. A positive outcome would motivate the need for de-identification algorithms to be applied to acceleration signals, embedded into wearable devices, in order to limit the unintentional release of personal details and ensure the necessary privacy by design and by default requirements.