DESI ONLINE SEMINAR
Abstract:
Wearable devices, such as fitness trackers, are increasingly popular and their number is expected to reach the billion by 2022. Although they bring substantial benefits for improving the quality of life of their users, they raise serious privacy issues. For instance, previous research demonstrated that fitness trackers data help infer sensitive information such as drug consumption. One particularly sensitive piece of personal information, which recently attracted substantial attention, is personality as it enables to influence individuals (e.g., voters in the Cambridge Analytica scandal).
In this work, we present the first empirical study on the inference of personality traits (w.r.t. the Big Five model) from fitness trackers data. We conduct an experiment with 200+ participants: we establish our ground truth by relying on the NEO-PI-3 questionnaire and collect the step count, heart rate, battery, activities, and sleep of the participants for four months. By following a principled machine-learning approach, we quantify our participants' personality privacy. Our results demonstrate that by using the seemingly innocuous data collected from fitness trackers, an adversary can improve the accuracy of the inference of extraversion and neuroticism, with statistical significance. We further study the importance of the different features (i.e., data types) on the accuracy of the inference.
Bio:
Noé Zufferey is a PhD student at UNIL-HEC Lausanne, working under the joint supervision of Prof. Kévin Huguenin, in the framework of the PrivateLife SNSF project. He started his PhD in early 2019. He earned his Msc in computer science from the BeNeFri network (Universities of Fribourg, Bern and Neuchâtel) with a specialization in logic.