PAFR: Privacy-Aware Friends Retrieval over Online Social Networks

Abstract

Online Social Networks (OSNs) are online services that people use to build social relations with other people. Friends retrieval over OSNs is an important activity among users. However, in existing friends retrieval solutions, user may leak private and sensitive information, such as personal data and friends relation. Traditional end-to-end encryption methods can protect users’ information, while it is not available for them to share contents with others. In this paper, aiming at preventing attack launched by the service provider, we propose PAFR, a Privacy-Aware Friends Retrieval scheme over OSNs. Inspired by private set interaction protocols, our scheme allows user to upload his encrypted contents to server, obliviously connect with others as friends, and retrieve friends with specific content. In addition, we design a dual-server model where two non-collude servers S1 and S2 perform secure sorting protocol to retrieve the ranked relevant friends without learning the underlying query and friends relationship. Compared with the approach with single server holding the whole knowledge of the OSN, each of them only has a part of the information make our scheme achieve a high level of privacy. Both security and performance analysis demonstrate that our scheme has both a very light user workload and a moderate server workload while being secure against user-server collusion.