Show me your app usage and I will tell who your close friends are: Predicting user's context from simple cellphone activity

Personal interactions using cell phones are so embedded in our everyday lives that they go almost unnoticed. We may try to protect ourselves from releasing sensitive information by increasing privacy protections, but how much can be inferred from our most basic phone usage? Using a largescale annotated dataset of cell phone usage, we build a predictor to determine location context (home, work, commute) and social relationships (with close friend, with family) based on the clock of the phone and sequences of apps executed. Surprisingly, we show that just using this basic information we can accurately predict whether someone is at home, at work, and/or with close friends, family. We note that this is almost inevitable because it only depends on using the phone and not the privacy settings. Our results suggest that our relationship with technology gives away more than we might suspect. This presents opportunities and challenges discussed in this paper.