Location Privacy Protection And Its Applications In Location-based Social Networking Services

With the popularity of smart portable devices, advances in wireless positioning technology, as well as the development of the mobile Internet, a new form of social networking services——the location-based social networking services (LBSNS) emerges and attracts a lot of people all over the world. It is the expansion and extension of the traditional social networking services under the mobile environment. Such kind of applications is established for the purpose of social interactions and contents sharing through mobile devices. It seamlessly integrates the physical space, information space and human activities together which makes social networking services no longer limited to a static environment, and closer to people’s daily life. With the development of LBSNS, it has to faces many challenges among which the location privacy protection problem has gained widespread concerns of researchers at home and abroad while the abuse of location information may lead to the leakage of personal privacy. A number of related efforts on location privacy protection in LBSNS have been conducted in the past few years. However, most of them are specified to certain scenarios. Furthermore, there is a lack of location privacy protection models concentrating on LBSNS, as well as researches on applications of protected locations to various fields of LBSNS.In this thesis, the location privacy protection issue in LBSNS is comprehensively analyzed. A generalized location privacy protection framework is proposed, as well as a location privacy protection model. Moreover, the usage of protected locations in regions of interest mining, friendship prediction, and location-based query in LBSNS is investigated.To summarize, the main contributions of this thesis are listed as follows:●The thesis analyzes the core issues such as why people want to share their locations, what are the benefits, and why do risks take place, etc. in LBSNS from the perspectives of user and developer respectively, taking the advantages of sociological and psychological theories. Based on the summarization of above work, the generalized location privacy protection framework (GLPPF) is proposed concentrating on the protection for the whole process of location information manipulations including exchange, storage, application and display.●The thesis introduces a modified k-anonymous spatial-temporal cloaking model (KSTCM). It makes a location record cannot distinguish from other k-1’s on the spatial, temporal and semantic annotation dimensions by generalization. An algorithm which helps to build the KSTCM model of location history with the minimum spatial and temporal range is proposed, as well as an algorithm to update real-time anonymized locations. Experimental results demonstrate that the location records can be protected efficiently by these algorithms.●The thesis explores the applications of protected locations to regions of interest (ROIs) mining and friendship prediction in LBSNS. For the ROIs mining, two kinds of ROIs, the hot regions and the personal regions are defined. Methods used to discover these regions mainly exploit the phenomenon that popular regions are frequently visited by different people leading to a high information entropy, while private regions are visited by few users, and thus with low entropy. On the other hand, a co-checkin relationship model based on KSTCM is introduced, and friendship can be accurately predicted by learning classifiers with analysis of diverse features. Experimental results show that protected locations perform well in both ROIs mining and friendship prediction which proves that location privacy protection and qualities of services can be both obtained in LBSNS.●The thesis studies the real-time location-based query in LBSNS with protected locations. It defines the scenario that a group of users prepare to gather in a place as the multi-object convergence (MOC) problem. In order to solve this kind of problem, an algorithm which utilizes the Voronoi graph and skyline query operator is proposed. In addition, with respect to the dynamic query conditions, two algorithms which are able to efficiently reduce the duplicate computing and considerably improve the performance of second-time queries are proposed.