Personalized location anonymity - A kernel density estimation approach

In recent years, the problem of location privacy protection in location-based service (LBS) has drawn a great deal of researchers’ attention. However, the existing technologies of location privacy protection rarely consider the personal visit probability and other side-information, which are likely to be exploited by attackers. In order to protect the users’ location privacy more effectively, we propose a Personal Location Anonymity (PLA) combining side-information to achieve k-anonymity. On the offline phase, we utilize Kernel Density Estimation (KDE) approach to obtain the personal visit probability for each cell of space according to a specific users’ visited locations. On the online phase, the dummy locations for each user’s query can be selected based on both the entropy of personal visit probability and the area of Cloaking Region (CR). We conduct extensive experiments on the real dataset to verify the performance of privacy protection degree, where the privacy properties are measured by the location information entropy and the area of CR.