A Differentially Private Matching Scheme For Pairing Similar Users Of Proximity Based Social Networking Applications

The pervasiveness of smartphones has made connecting with users through proximity based mobile social networks commonplace in today's culture. Many such networks connect users by matching them based on shared interests. With ever-increasing concern for privacy, users are wary of openly sharing personal information with strangers. Several methods have addressed this privacy concern such as encryption and k-anonymity, but none address issues of eliminating third party matches, achieving relevant matches, and prohibiting malicious users from inferring information based on their input into the system. In this paper, we propose a matching scheme that accurately pairs similar users while simultaneously providing protection from malicious users inferring information. Specifically, we match users in a proximity-based social network setting adapted from a framework of differential privacy. This eliminates the need for third-party matching schemes, allows for accurate matching, and ensures malicious users will be unable to infer information from matching results.