Hong kyu Lee

I am a PhD candidate at Emory University. I am a member of AIMS Lab, under the supervision of Prof. Li Xiong and Prof. Carl Yang. My research is dedicated to advancing the Privacy and Security of AI and develop Trustworthy AI models. My primary focus is on data privacy and the security of machine learning models. My projects involve developing robust machine unlearning mechanisms and differentially private training, and probing model vulnerabilities through membership inference attacks. I have experience implementing these methods on complex architectures, including large language models (LLMs), vision transformers, and multimodal models.

A central question driving my work is how to precisely define the “learned state” of a model. I believe that a deeper understanding of this concept is essential for accurately assessing privacy risks and, consequently, for building systems that can truly and verifiably “unlearn” information.