A dummy-based location privacy protection scheme with semantic correlation of moving paths

Dummy-based methods protect users’ location privacy by requiring them to submit their real locations along with a large number of dummies to location service providers (LSPs). However, existing solutions do not consider the spatio-temporal and semantic correlations between location sequences from an overall query perspective when generating dummies. This leads that LSPs can identify dummies by analyzing the relationships between location sequences in the resulting movement paths, and may even directly infer users’ real locations. Therefore, existing dummy-based schemes are insufficient in effectively preserving the users’ location privacy. To address this issue, this paper utilizes neural networks to learn the semantic and spatio-temporal features between locations embedded in users’ real movement paths, and integrates the transition probabilities across location sequences to quantify the influence of query timing on these features. Based on this, a dummy-based location privacy protection scheme with semantic correlation of movement paths is proposed. Theoretical analysis and extensive experiments show that the proposed scheme can efficiently help users generate dummy sequences that are highly indistinguishable from the real ones in terms of both spatio-temporal and semantic features, thereby effectively protecting the users’ location privacy.