Document Type
Dissertation
Date of Award
5-31-2023
Degree Name
Doctor of Philosophy in Information Systems - (Ph.D.)
Department
Informatics
First Advisor
Hai Nhat Phan
Second Advisor
Yi-Fang Brook Wu
Third Advisor
Yi Chen
Fourth Advisor
Tong Sun
Fifth Advisor
Xiong Li
Abstract
Nowadays, machine learning (ML) becomes ubiquitous and it is transforming society. However, there are still many incidents caused by ML-based systems when ML is deployed in real-world scenarios. Therefore, to allow wide adoption of ML in the real world, especially in critical applications such as healthcare, finance, etc., it is crucial to develop ML models that are not only accurate but also trustworthy (e.g., explainable, privacy-preserving, secure, and robust). Achieving trustworthy ML with different machine learning paradigms (e.g., deep learning, centralized learning, federated learning, etc.), and application domains (e.g., computer vision, natural language, human study, malware systems, etc.) is challenging, given the complicated trade-off among utility, scalability, privacy, explainability, and security. To bring trustworthy ML to real-world adoption with the trust of communities, this study makes a contribution of introducing a series of novel privacy-preserving mechanisms in which the trade-off between model utility and trustworthiness is optimized in different application domains, including natural language models, federated learning with human and mobile sensing applications, image classification, and explainable AI. The proposed mechanisms reach deployment levels of commercialized systems in real-world trials while providing trustworthiness with marginal utility drops and rigorous theoretical guarantees. The developed solutions enable safe, efficient, and practical analyses of rich and diverse user-generated data in many application domains.
Recommended Citation
Lai, Thi Kim Phung, "Trustworthy machine learning through the lens of privacy and security" (2023). Dissertations. 1664.
https://digitalcommons.njit.edu/dissertations/1664