Inferring social ties in pervasive networks: an on-campus comparative study

WiFi base stations are increasingly deployed in both public spaces and private companies, and the increase in their density poses a significant threat to the privacy of users. Prior studies have shown that it is possible to infer the social ties between users from their (co-)location traces but they lack one important component: the comparison of the inference accuracy between an internal attacker (e.g., a curious application running on the device) and a realistic external eavesdropper (e.g., a network of sniffing stations) in the same field trial. We experimentally show that such an eavesdropper can infer the type of social ties between mobile users better than an internal attacker.