About
I am a postdoctoral researcher at the University of Waterloo in the Secure Systems Group (SSG), part of CrySP within the David R. Cheriton School of Computer Science, where I work with Prof. N. Asokan on AI security and privacy. Before that, I obtained my PhD degree in applied mathematics at the Academy of Mathematics and Systems Science, Chinese Academy of Sciences, under the supervision of Prof. Xiao-Shan Gao.
News
- [2027.01] I will be teaching CS 453: Computer and Systems Security at the University of Waterloo.
- [2026.10] Invited to serve as an Area Chair for ICML 2027.
- [2026.10] Invited to serve as Proceedings Chair for Canadian AI 2027.
- [2026.09] Invited to serve as an Area Chair for ICLR 2027.
- [2026.09] Our paper Extracting Training Data from Diffusion Language Models via Infilling is accepted to NeurIPS 2026.
More news
- [2026.08] Our paper Cross-Session Decomposition Attacks is accepted to EMNLP 2026 Findings.
- [2026.05] Co-organized the Data and Model Protection workshop at Canadian AI 2026.
- [2026.02] Our paper Towards Robust Content Watermarking Against Removal and Forgery Attacks is accepted to CVPR 2026 Findings.
- [2025.09] Our paper BridgePure is accepted to NeurIPS 2025.
- [2025.09] Joined the CrySP group at the University of Waterloo as a postdoctoral researcher, working with Prof. N. Asokan.
Research
My research focuses on trustworthy machine learning, with an emphasis on AI security, privacy, and copyright protection. I study how machine learning systems, from image models to large language models and AI agents, can be attacked, and how to build principled, deployable defenses that hold up under realistic adversaries. My current topics include:
- Copyright and IP protection: unlearnable examples, detecting unauthorized training use, and watermarking AI-generated content.
- Privacy and memorization: training-data extraction from (diffusion) language models and machine unlearning.
- Security of LLMs and AI agents: prompt injection, jailbreaking, and data poisoning in LLM post-training.
- Theoretical foundations: adversarial robustness, generalization and stability, and bi-level optimization.
Looking ahead, I am especially interested in copyright protection for the age of AI agents: giving rights holders technical means to verify how AI systems use their work, control that use, and attribute AI-generated content to its true source.
Selected Publications
See all publications. * denotes equal contribution.
-
Extracting Training Data from Diffusion Language Models via Infilling.
NeurIPS 2026.Shows that diffusion language models leak more training data than previously believed, by exploiting their ability to fill in text.
-
Towards Robust Content Watermarking Against Removal and Forgery Attacks.
CVPR 2026 Findings.Watermarks for AI-generated images that resist both removal and forgery attacks.
-
BridgePure: Limited Protection Leakage Can Break Black-Box Data Protection.
NeurIPS 2025.Shows that a small amount of leaked protected data is enough to break black-box data protection.
-
Efficient Availability Attacks against Supervised and Contrastive Learning Simultaneously.
NeurIPS 2024.Efficient unlearnable examples that protect data against both supervised and contrastive learning.
Service & Teaching
- Area Chair: ICLR 2027, ICML 2027.
- Organizing: Proceedings Chair, Canadian AI 2027; Co-organizer, Data and Model Protection Workshop at Canadian AI 2026.
- Reviewer: NeurIPS, ICML, ICLR, AAAI, ACL Rolling Review (Top Reviewer recognition multiple times).
- Teaching: Instructor, CS 453 Computer and Systems Security (Winter 2027); Invited expert, CS 858 Trustworthy Machine Learning (Fall 2025, Fall 2026), University of Waterloo.
- Mentoring: Worked closely with junior graduate students, resulting in co-authored papers at top-tier conferences; supervised four undergraduate research projects (URA) at the University of Waterloo.
Misc
I'm open to research discussion and collaborations on any related topics. Feel free to email me to schedule a meeting when I'm available in my calendar.
Outside of research, I enjoy playing badminton and cooking.