Fully Anonymous Profile Matching in Mobile Social Networks
|Name||Fully Anonymous Profile Matching in Mobile Social Networks|
In this paper, we study user profile matching with privacy-preservation in mobile social networks (MSNs) and introduce a family of novel profile matching protocols. We first propose an explicit Comparison-based Profile Matching protocol (eCPM) which runs between two parties, an initiator and a responder. The eCPM enables the initiator to obtain the comparison-based matching result about a specified attribute in their profiles, while preventing their attribute values from disclosure.We then propose an implicit Comparison-based Profile Matching protocol (iCPM) which allows the initiator to directly obtain some messages instead of the comparison result from the responder. The messages unrelated to user profile can be divided into multiple categories by the responder. The initiator implicitly chooses the interested category which is unknown to the responder. Two messages in each category are prepared by the responder, and only one message can be obtained by the initiator according to the comparison result on a single attribute. We further generalize the iCPM to an implicit Predicate-based Profile Matching protocol (iPPM) which allows complex comparison criteria spanning multiple attributes. The anonymity analysis shows all these protocols achieve the confidentiality of user profiles. In addition, the eCPM reveals the comparison result to the initiator and provides only conditional anonymity; the iCPM and the iPPM do not reveal the result at all and provide full anonymity. We analyze the communication overhead and the anonymity strength of the protocols. We then present an enhanced version of the eCPM, called eCPM+, by combining the eCPM with a novel prediction-based adaptive pseudonym change strategy. The performance of the eCPM and the eCPM+ are comparatively studied through extensive trace-based simulations. Simulation results demonstrate that the eCPM+ achieves significantly higher anonymity strength with slightly larger number of pseudonyms than the eCPM.
|ieee paper year||2013|