Anonymization of geosocial network data by the (k, l)-degree method with location entropy edge selection

Abstract

Geosocial networks (GSNs) have become an important branch of location-based services since sharing information among friends is the additional feature to provide information based on the user's current location. The growing popularity of location-based services contribute to the development of highly customized and flexible utilities. However, providing customized services relates to collecting and storing a large amount of users' information. In this paper, we focus on the privacy-preserving concern in publishing GSN datasets. We introduce a new (k, l)-degree anonymization method to prevent the re-identification attack in the published GSN dataset. The presented method anonymizes users' social relationships as well as location-based information in GSN. We propose the new (k, l)-degree anonymization algorithm which modifies the network structure with a sequence of edge editing operations. GSN is newly represented by the combination of social network describing social ties between users and affiliation network linking users with their checked-in locations. Furthermore, we innovatively use the location entropy metric in the proposed GSN anonymization method. The location entropy measures the importance of the visited locations in the edge selection procedure of the (k, l)-degree anonymization algorithm. We explore the usability of the algorithm by running experiments on real-world geosocial network datasets, Gowalla and Brightkite.