Privacy-preserving spatial keyword location-to-trajectory matching

Ning Wang,Jianping Zeng,Wenxing Hong,Shunzhi Zhu

DOI: https://doi.org/10.1007/s10619-020-07290-2

IF: 0.974

2020-04-24

Distributed and Parallel Databases

Abstract:Location-to-trajectory matching plays an important role in trajectory data management and analysis. In this paper, we propose and study a novel problem of privacy-preserving spatial keyword location-to-trajectory matching (PPSKLT Matching). Given a set O of locations with keywords, a set T of activity trajectories, a spatial matching threshold \(\theta _{\text{sp}}\), and a keyword matching threshold \(\theta _{\text{kw}}\), the PPSKLT matching finds all location-trajectory pairs from O and T while preserving the users' privacy. We believe that the PPSKLT matching may benefit many mobile applications such as trajectory activity analysis, event tracking, and so on. The PPSKLT matching is challenging due to three reasons: (1) how to define the spatial keyword similarity measure between locations and trajectories, (2) how to prune the search space effectively, and (3) how to preserve the users' privacy during query processing. To overcome these challenges and address the problem efficiently, we develop a novel network expansion algorithm (NEA). We define a pair of upper and lower bounds on the spatio-textual similarity to prune the search space. We also define a privacy-preserving mechanism to protect users' privacy. We conduct extensive experiments on spatio-textual data sets to verify the performance of the developed algorithms.

computer science, information systems, theory & methods