Feel Free to Check-in: Privacy Alert against Hidden Location Inference Attacks in GeoSNs

摘要

Check-in services, one of the most popular services in Geo-Social Networks (GeoSNs) may cause users' personal location privacy leakage. Although users may avoid checking in places which they regard as sensitive, adversaries can still infer where a user has been through linkage of multiple background information. In this paper, we propose a new location privacy attack in GeoSNs, called hidden location inference attack, in which adversaries infer users' location based on users' check-in history as well as check-in history of her friends and similar users. Then we develop three inference models (baseline inference model, CF-based inference model and HMM-based inference model) to capture the hidden location privacy leakage probability. Moreover, we design a privacy alert framework to warn users the most probable leaked locations. At last, we conduct a comprehensive performance evaluation using two real-world datasets collected from Gowalla and Brightkite. Experiment results show the accuracy of our proposed inference models and the effectiveness of the privacy alert framework.