Privacy-Preserving Dense Subgraph Discovery in Mobile Social Networks

With the ubiquity of social networks and mobile devices, various new mobile applications for impromptu group formation and communications have emerged recently, such as Meetup, Plancast, Yahoo! Upcoming, Eventbrite and Epinions. Among those important applications, finding a dense subgraph in a social network plays a crucial role, but the current algorithms are only designed for online social networks. Different from online social networks, no central repository is available in mobile social networks (MSNs). Finding a dense subgraph in a distributed manner without compromising the individual privacy is very challenging, because a dense subgraph usually includes abundant friendship information. To address this important need for MSNs, we propose a privacy-preserving k-core discovery protocol, named PKcore. We show that the solution group returned by our protocol is optimal. In addition, we guarantee that the MSN users in our protocol do not reveal any social links to both the initiator and the other contacts. We also prove that the privacy properties based on Canetti´s model always hold. Finally, we analyze the computation and communication overheads and show that our protocol is feasible for current mobile devices.