Exploring Differential Privacy in Geospatial Demographics: Algorithms, Challenges, and Insights from Qatar

Abstract

As the demand for data-driven insights grows, protecting privacy in demographic datasets, especially those with geospatial information, is becoming critical. Differential Privacy (DP) provides a robust approach by adding controlled noise to data, reducing the risk of re- identification. This thesis examines two DP algorithms — the Uniform Grid (UG) and Adaptive Grid (AG) — and proposes an enhancement through the Modified Adaptive Grid (m-AG), which is optimized to handle areas with sparse data more effectively. Using real-world population data from Qatar, this study evaluates the performance of UG, AG, and m-AG in terms of their privacy-utility trade-offs, focusing on computational efficiency and data accuracy. The findings indicate that m-AG reduces cumulative error and runtime by adapting noise application based on data density, demonstrating its suitability for applications where geospatial data privacy is crucial. This research contributes to the ongoing development of privacy-preserving techniques in the geospatial domain, offering a practical improvement for balancing data utility and privacy protection.

Date of Award	2024
Original language	American English
Awarding Institution	HBKU College of Science and Engineering