Spatio-Temporal Features Based Sensitive Relationship Protection in Social Networks

Xiangyu Liu, Mandi Li, Xiufeng Xia, Jiajia Li, Chuanyu Zong, Rui Zhu

Abstract: Without considering the spatio-temporal features of check-in data, there exists privacy leakage risk for the sensitive relationships in social networks. In this paper, a Spatio-Temporal features based Graph Anonymization algorithm (denoted as STGA) is proposed, in order to protect sensitive relationships in social networks. We firstly devise a relationship inference algorithm based on users’ neighborhoods and spatio-temporal features of check-in data. Then, in STGA, we propose an anonymizing method through suppressing the edges or check-in data to prevent the inference of sensitive relationships. We adopt a heuristic to obtain the inference secure graph with high data utility. A series of optimizing strategies are designed for the process of edge addition and graph updating, which reduce the information loss meanwhile improving the efficiency of the algorithm. Extensive experiments on real datasets demonstrate the practicality, high data utility and efficiency of our methods.