Synthesizing Privacy Preserving Traces: Enhancing Plausibility With Social Networks

Due to the popularity of mobile computing and mobile sensing, users’ traces can now be readily collected to enhance applications’ performance. However, users’ location privacy may be disclosed to the untrusted data aggregator that collects users’ traces. Cloaking users’ traces with synthetic traces is a prevalent technique to protect location privacy. But the existing work that synthesizes traces suffers from the social relationship based de-anonymization attacks. To this end, we propose <inline-formula> <tex-math notation="LaTeX">$W^{3}{-}tess$ </tex-math></inline-formula> that synthesizes privacy-preserving traces via enhancing the <italic>plausibility</italic> of synthetic traces with social networks. The main idea of <inline-formula> <tex-math notation="LaTeX">$W^{3}{-}tess$ </tex-math></inline-formula> is to credibly imitate the temporal, spatial, and social behavior of users’ mobility, sample the traces that exhibit similar three-dimension mobility behavior, and synthesize traces using the sampled locations. By doing so, <inline-formula> <tex-math notation="LaTeX">$W^{3}{-}tess$ </tex-math></inline-formula> can provide “<italic>differential privacy</italic>” on location privacy preservation. In addition, compared to the existing work, <inline-formula> <tex-math notation="LaTeX">$W^{3}{-}tess$ </tex-math></inline-formula> offers several salient features. First, both location privacy preservation and data utility guarantees are theoretically provable. Second, it is applicable to most geo-data analysis tasks performed by the data aggregator. Experiments on two real-world datasets, loc-Gwalla and loc-Brightkite, have demonstrated the effectiveness and efficiency of <inline-formula> <tex-math notation="LaTeX">$W^{3}{-}tess$ </tex-math></inline-formula>.