Woohyuk Choi (최우혁)
I am a Ph.D. student in CompSec Lab, Department of Electrical and Computer Engineering, Seoul National University. I am advised by Prof. Byoungyoung Lee.
I am interested in security for AI and system security in general. In particular, my research focuses on AI agent security. I designed Prompt Flow Integrity and DualView, a system-level mitigation against prompt injection attacks in LLM agents. Also I discovered Agent Data Injection (ADI), a new class of attacks that exploits the misinterpretation of untrusted data as trusted data in AI agents.
I am currently seeking internship or visiting researcher opportunities in the areas of system security and AI agent security. If you are interested in collaborating or have opportunities available, please feel free to reach out to me at 00cwooh@snu.ac.kr.
Publications
Agent Data Injection Attacks are Realistic Threats to AI Agents
Woohyuk Choi*, Juhee Kim*, Taehyun Kang, Jihyeon Jeong, Luyi Xing, and Byoungyoung Lee
arXiv:2607.05120, 2026 (* Equal contribution)
DualView: Preventing Indirect Prompt Injection in Personal AI Agents
Woohyuk Choi*, Juhee Kim*, Taehyun Kang, Youngmin Kim, and Byoungyoung Lee
arXiv:2607.03821, 2026 (* Equal contribution)
GHost in the SHELL: A GPU-to-Host Memory Attack and Its Mitigation
Sihyun Roh, Woohyuk Choi, Jaeyoung Chung, Yoochan Lee, Suhwan Song and Byoungyoung Lee
IEEE Symposium on Security and Privacy (SP) 2026 (accepted, to appear)
Prompt Flow Integrity to Prevent Privilege Escalation in LLM Agents
Woohyuk Choi*, Juhee Kim*, and Byoungyoung Lee
arXiv:2503.15547, 2025 (* Equal contribution)
Experience
Internship, Samsung MX (Mobile eXperience), Security Engineering Group
Suwon, Korea, Jan 2024 - Feb 2024