Abstract
dc:description.abstract<p>Modern artificial intelligence (AI) systems have transformed critical domains such as healthcare, software engineering, finance, and the legal system. Despite their broad impact, concerns about trustworthiness, especially regarding privacy and security, remain major obstacles to wider adoption. Addressing these concerns requires both a systematic understanding of the privacy and security risks inherent in AI systems and the development of principled foundations for trustworthy AI that safeguard client privacy and security. This goal is particularly challenging because of the complexity of modern large-scale AI systems, the trade-offs between privacy and model utility, and the need to simultaneously ensure other important properties such as fairness and robustness in the AI decision-making process.</p><p>This dissertation addresses these challenges by studying the privacy and security risks of modern large-scale AI systems and by developing optimized mechanisms with theoretical guarantees that mitigate such risks while preserving strong downstream performance across diverse domains, including computer vision, natural language processing, and software engineering. The dissertation is structured around two closely connected research thrusts. The first focuses on understanding and mitigating privacy risks in large-scale and relational AI systems through methods that provide strong privacy guarantees while maintaining high task utility. The second focuses on leveraging AI to detect and mitigate security risks, with the broader goal of improving the safety of digital systems, including AI systems themselves.</p><p>Across both directions, the proposed methods are supported by extensive theoretical analysis and empirical evaluation. The theoretical results establish rigorous guarantees for the developed mechanisms, while the experimental studies demonstrate their effectiveness and practicality in real-world settings. Together, these contributions advance the foundations of trustworthy AI by showing how privacy and security can be strengthened without undermining the performance needed for deployment. More broadly, this dissertation helps enable the responsible adoption of modern large-scale AI systems in high-stakes domains where trust is essential.</p>
Degree
thesis:*- Name thesis:degree_name
- Doctor of Philosophy in Data Science - (Ph.D.)
- Discipline thesis:degree_discipline
- Data Science
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Tran, Khang
- 0000-0002-5154-8119
- Contributors dc:contributor
-
- Hai Nhat Phan
- Cristian Borcea
- Yi Chen
Subjects
dc:subject × 8Identifiers
dc:identifier.*- Repository record dc:identifier
- https://digitalcommons.njit.edu/dissertations/1890