I study security and privacy risks in data-driven and AI systems, from discovering new vulnerabilities to auditing and controlling them. My current research focuses on three directions:
ManuscriptRATION: Towards Quantitative Risk-Based Access Control for AI Agents Yuntao Du, Harsh Chaudhari, Yuetian Chen, Ron Heichman, Ashish Jayamohan, Kumar Pallav, Mahim Patel, Kaiyuan Zhang, Kyle Polley, Jerry Ma, Ninghui Li Manuscript
AI Security & SafetyDefenseA new security framework for AI agents.
ACL 2026 (Findings)Automated Profile Inference with Language Model Agents Yuntao Du, Zitao Li, Bolin Ding, Yaliang Li, Hanshen Xiao, Jingren Zhou, Ninghui Li Findings of the Association for Computational Linguistics
AI Security & SafetyHackingFirst study showing that LLM agents enable automated doxing at web scale.
NDSS 2026Cascading and Proxy Membership Inference Attacks Yuntao Du, Jiacheng Li, Yuetian Chen, Kaiyuan Zhang, Zhizhen Yuan, Hanshen Xiao, Bruno Ribeiro, Ninghui Li Network and Distributed System Security Symposium
Machine Learning PrivacyFormulates and categorizes MIAs and first exploits membership dependencies.
USENIX Security 2026Imitative Membership Inference Attack Yuntao Du, Yuetian Chen, Hanshen Xiao, Bruno Ribeiro, Ninghui Li USENIX Security Symposium
Machine Learning PrivacyA new shadow training paradigm for MIAs with significantly reduced computation.
ACM CCS 2026Privacy Leakage from a Thousand Words: Millipixel Location Recovery from Dot Maps Yuntao Du⭑, Tanishq Pauskar⭑, Hao Wang, Jing Su, Ninghui Li ACM Conference on Computer and Communications Security
Data PrivacyHackingAchieve ~1-meter attack accuracy in location recovery from national-scale dot maps.
Impact This work informed QGIS's privacy guidance. Working with the U.S. CDC to revise its cartographic guidelines.
ACM CCS 2025Systematic Assessment of Tabular Data Synthesis Yuntao Du, Ninghui Li ACM Conference on Computer and Communications Security
Data PrivacyProposes a unified evaluation framework for tabular data synthesis algorithms.