sunpill kim

Sunpill Kim

김선필 · 金善必

Postdoctoral researcher at Hanyang University, working at the intersection of applied mathematics, security, and AI — using the geometry of learned representations to study what modern recognition systems leak, and how to defend them.

Sunpill Kim
About

I received my B.S. (2020) and Ph.D. (2026) in Mathematics from Hanyang University, and have been a member of the Cryptology & Algorithm Lab under Professor Jae Hong Seo since 2018. I am currently a postdoctoral researcher at the same group.

From January 2023 to January 2024 I was an ARAP scholar at the A*STAR Institute for Infocomm Research (Singapore), working with Dr. Khin Mi Mi Aung and Dr. Yong Kiam Tan.

My work studies the security of metric-learning–based recognition systems — face recognizers, speaker recognizers, and the vision–language models built on the same geometric primitives. The thread that runs through it is mathematical: representation spaces have structure, and that structure governs both what an attacker can extract and what a defender can hide. The pieces I have worked on so far are non-adaptive attacks, biometric template protection, and certified robustness for recognition. See the research page for an overview, or publications for the full list.

I have served as a reviewer for NeurIPS, CVPR (2024–26), ECCV, BMVC, and IEEE journals (TIFS, TDSC). My doctoral work was recognized with Hanyang's Outstanding Ph.D. Dissertation Award (2026) and supported by the 1st Graduate Presidential Science Scholarship (2024–26, sole finalist in Mathematics). I led an NRF-funded project on secure biometric authentication as Principal Investigator (2024–25).

Recent

all news