Privacy implications of geosocial proximity

Geosocial networks, social networks integrating their users’ location, meet an undeniable success. While traditional social networks are already subject to privacy breaches, the addition of location make such breaches to pile up. However, a not obvious question is : does this combination create a new type breaches which were not previously conceivable by the sole use of social networks or location information? We address this matter in two experiments. A first one which takes geosocial network users’ profiles and classify them depending on whether their friends live globally close to or far from them. We find that 70 % of users in our Foursquare dataset have their friends living relatively close to them. The second experiment that we led use the check-ins of Gowalla and Brightkite users to discover their social graphs. The check-ins model the users’ movements then we assume that users whose modellings are sufficiently similar are friends. This assumption comes from the fact that one often makes new friends in the places that he likes and where he spends a long time. Our partial findings seems to indicate that we are able to discover a half of a user’s social graph in average. Thus, we attempt a beginning of answer saying that geosocial networks actually allow new types of attacks against their users’ privacy.