Check-Ins And Photos: Spatiotemporal Correlation-Based Location Inference Attack And Defense In Location-Based Social Networks

In Location-Based Social Networks (LBSNs) users' interaction is largely carried through check-ins and photo sharing. The study of location privacy issues with check-ins has yielded different Location Privacy-Preserving Mechanisms (LPPMs), including dummies to hide user's whereabouts in a set of dummy locations without using a trusted-third party(TTP). However, the impact of shared photos on location privacy is yet to be understood. Our experiment on real data from a LBSN reveals that, like the check-ins, the spatial distribution of shared photos can influence user's location privacy. In this paper, we propose an inference model based on spatial distribution of historical check-ins and photos, and show that it is possible to deduce user's location at a high accuracy through spatiotemporal analysis of multiple events, comprising check-ins and photos. In the process, we evaluate state-of-the-art dummy mechanisms with the proposed inference model to stress the limitation of existing LPPMs. Then, we design a LPPM, called photo-check, to protect user privacy in LBSN for both check-in and photos; and carry out the experiment with real data from Foursquare to show its effectiveness and efficacy.