Privacy Issues Of Proximity-based Matching In Mobile Social Network

With the explosive increase of mobile devices, applications in Proximity-based Mobile Social Networks(PMSN) have been widely developed and research topics about them have also attracted great attentions. However, in the process of similarity matching, privacy preserving of users' profile items and improvement of matching efficiency are both urgent requirements. In the matching schemes based on private set intersection, such as Garbled Bloom Filter(GBF), both security and efficiency need improvement. In the profile matching schemes based on similarity, there also exist some problems such as lack of difference on profile items, limited expressive ability of similarity metrics, and high time cost of profile matching.Aiming at solving the abovementioned problems in Proximity-based Mobile Social Networks, our research mainly focuses on the following aspects:(1) Based on analysis of the existing GBF scheme, we observe its limitation in security and communication overhead and propose a series of enhanced GBF schemes to improve the two aspects. In the scheme of avoiding exposing set element in collision, we prevent the adversary from confirming a collision such that the set element of user is protected. In the scheme of avoiding dictionary attack, we involve keyed Message Authentication Code(MAC) and similar authentication methods to defend dictionary attack, since the adversary is unable to verify whether his attack is success. And the introduced computation and communication overhead are both limited. In the scheme of reducing communication overhead, we add shift to the original location of set element in GBF, which makes its location is no longer stable hence improves security. And based on the improved security, we shorten the string length in GBF to reduce the communication overhead. All the proposed schemes can be used in the later privacy-preserving profile matching scheme to protect item name from exposed.(2) Based on the analysis of the existing fine-grained profile matching scheme, we observe that its expressive ability and communication efficiency both need enhancement. We propose a user self-controllable profile matching protocol in privacy-preserving mobile social networks. By introducing the weighted Manhattan distance as similarity metrics, where the weights and the threshold are all chosen by the users themselves, users can customize the matching metrics to involve their own matching preference and to make the matching results more precise. By utilizing the abovementioned enhanced GBF schemes, our scheme can also protect the privacy of users' profile item names. Moreover, extensive performance evaluations are conducted to illustrate that our scheme is more efficient than a relevant protocol in terms of computation and communication overhead, especially when the maximum value of profile item is large.We focus on the privacy-preserving problems in Proximity-based Mobile Social Networks and propose a series of solutions which improve both security and efficiency of existing schemes and make profile matching more applicable in terms of security, reliability and efficiency.