Differentially-Private Analytics for Qatar Traffic Data

Abstract

As artificial intelligence (AI) systems continue to evolve, striking a balance between their growing data requirements and the need to maintain privacy and security has become a critical research focus. In the context of Qatar, AI applications are increasingly utilized in various sectors, such as transportation and healthcare, bringing about significant improve- ments in efficiency and effectiveness. Ensuring the privacy of sensitive information in these domains is important as the utilization of data plays a significant role in driving these inno- vations. This thesis investigates a technique for releasing real geo-tagged traffic data from Qatar with Differential Privacy (DP), aiming to protect sensitive information while supporting AI analysis. The research explores the principles of DP and various DP-compliant mecha- nisms, techniques for aligning the real-world dataset with the simulated dataset, evaluating the feasibility of various sanitization algorithms for the Qatar dataset, and understanding their trade-offs. Subsequently, these findings demonstrate the effectiveness of applying DP on Qatar traffic data, not only preserving data privacy and quality but also facilitating AI advancements in the region. With a growing emphasis on data security and privacy, the adoption of DP principles and DP-compliant methods will be essential in fostering trust and encouraging the widespread use of AI in various sectors, driving progress and innova- tion.

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