Clemson University researchers lead an international team to understand and support end-user Smart Home privacy management.
Smart Home devices like smart lights, thermostats, and doorbells are gaining popularity because of the connected and automated experience they render to their users. This user experience is made possible by collecting and processing data about end-user behaviors. As this data is deemed sensitive and intimate, users tend to demand fine-grained control over their privacy preferences. But are users equipped to exert such a detailed level of control over their Smart Home devices?
HATLab researchers Dr. Bart P. Knijnenburg, Ph.D. student Paritosh Bahirat and Dr. Yangyang He (Alumnus) joined hands with Dr. Martijn Willemsen and Dr. Qizhang (Kevin) Sun from TU Eindhoven’s (Netherlands) Process Tracing Lab to answer this very question. They found that while Smart Home users’ privacy decision should ideally depend on their cognitive evaluation of contextual factors (e.g. the type of device, type of data, and recipient), subtle design choices such as the default setting and the framing of a privacy setting can have a detrimental effect on users’ privacy decisions, causing them to ignore the context and make sub-optimal privacy decisions.
The HATLab researchers also worked on a solution to this problem: using machine learning techniques they were able to cluster Smart Home users’ privacy decisions into a set of five comprehensive privacy profiles. Asking end-users to choose one of these profiles before delving into the more detailed preferences circumvents the detrimental effects of defaults and framing, and significantly simplifies users’ privacy management tasks.
Work on this project was published in ACM and IEEE journal articles received a Best Paper Award at the 2018 ACM Conference on Intelligent User Interfaces (IUI) and will be presented at the upcoming ACM Conference on Human Factors in Computing Systems (CHI).