| 研究実績の概要 |
During FY 2022, we continued to explore both privacy and utility issues arising from Differential Privacy (DP) in the context of spatiotemporal correlations. Regarding privacy issues, we demonstrated that road networks could expose vulnerabilities in users' location data, even under the protection of DP. Essentially, attackers can exploit prior knowledge about road networks to deduce true locations from noisy (perturbed) locations. For utility issues, we designed post-processing approaches that leverage spatiotemporal correlations as prior information. The idea is to treat correlations as a property of the data, allowing us to model post-processing as an optimization problem constrained by data correlations. Our method significantly improved the utility of privacy-protected data.
|
