Computing and Information Systems - Theses
Permanent URI for this collection
Search Results
1 - 1 of 1
-
ItemPrivately counting trajectories, regions and tuning spatial data structures( 2017)Location data is used widely, from ride-sharing applications in consumer mobile to traffic management in urban planning. Mining of spatial data is an enabling technology for mobile services, Internet-connected cars, and the Internet of things. However, the very distinctiveness of spatial data that drives utility, can come at the cost of user privacy. In this thesis, we propose mechanisms for preserving user location privacy while facilitating third party spatial data analytics for applications such as facility location services and traffic planning. First, aggregation as a qualitative privacy approach is utilised, controlling utility by proposing the Connection Aware Spatial Euler (CASE) histogram for processing point trajectories into a data structure suitable for responding to range queries. In addition, mechanisms are also developed under the strong guarantee of differential privacy, not only for interactive spatial histogram release but also for the non-interactive setting. Secondly, we propose a non-interactive differentially-private approach to counting planar bodies representative of users’ spatial regions. We formulate novel constrained inference to improve the utility-privacy trade-off for range queries. And finally, we propose a first end-to-end differentially-private mechanism for releasing parameter-tuned spatial data structures. Our mechanism E2EPriv leverages a general-purpose approach to histogram parameter tuning via privately-optimising data-dependent bounds on error. Comprehensive experimental results on datasets of a range of scales, levels of sparsity and uniformity, establish that our mechanisms achieve competitive utility while preserving privacy.