Key Techniques For Privacy Preserving On Location-based Social Networks

With the booming of wireless sensor network and location techniques, more and more users have started to use Location-based Social Network (LBSN) applications. Therefore, a large amount of LBSN data is produced. LBSN data is so important to infer the behavior of users, to explore the similarity between users, to recommend friends or interesting locations, and so on. However, the direct publishing of LBSN data may bring a great threat to the user’s privacy information. Moreover, the anonymity of LBSN data has not been studied. This thesis mainly studies privacy protection technology of LBSN data.At present, although there are a mass of social network privacy protection technologies and location anontmity technologies, these technologies are not applicable to LBSN data. LBSN data not only contains the relationship between users of social network, but also contiins the relationship between the users and locations, which makes LBSN data anonymity technology more challenging.This thesis has made a systematic review on social network privacy protection technologies and location anontmity technologies. In view of different background knowledge attackers may get, this thesis proposes two privacy disclosure issues on LBSN, including the problem of nodes identity leakage and sensitive relation leakage, and the problem of nodes identity leakage caused by user’s rank sequence.To prevent nodes identity reveal and sensitive relation reveal, an anonymity model called k-link safe grouping has been presented in this thesis. We also propose two algorithms to implement the anonymity model, called k-LSG algorithm and k-ILSG algorithm. k-LSG algorithm first groups the location nodes using k-safe grouping, and then handles the out-link group and in-link group produced in grouping process. Furthermore, to improve the utility of anonymous data, we propose the algorithm with high utility, which is called k-ILSG. Secondly, in order to prevent an attacker from recognizing the identification of nodes according to the rank sequence, k-rank anonymity model for LBSN graph has been proposed in this thesis. We have also proposed a heuristic algorithm RA to obtain anonymous rank sequences and a reconstruction algorithm RRL to obtain anonymous LBSN graph respectively.We have proved the security of k-link safe grouping anonymity methods and k-rank anonymity method, and the high utility of anonymous data with a lot of experiment on real datasets.