Xun Wang

About Me

I am a third-year Ph.D. student at CISPA Helmholtz Center for Information Security. I am supervised by Prof. Michael Backes, with Prof. Franziska Boenisch as my co-advisor and Prof. Adam Dziedzic as my mentor. I work with Franziska and Adam daily at SprintML Lab.

Prior to my doctoral studies, I obtained my M.S. degree in Computer Science from Technical University of Munich (TUM) in 2023, graduating with high distinction. I completed my B.S. degree at Tongji University in 2020.

Research Interests

My research focuses on trustworthy and efficient machine learning, with an emphasis on how foundation models are adapted, deployed, and audited. My interests include:

• Model Provenance and Ownership: Auditing the identity of models served through agentic APIs and developing fingerprints to verify ownership of generative models.
• Privacy-Preserving Learning and Auditing: Designing private adaptations and soft prompt transfer methods for LLMs, and measuring empirical privacy risks through membership inference and data extraction attacks.
• Efficient Machine Learning: Developing architecture-agnostic structured pruning and efficient knowledge transfer methods that reduce the cost of adapting and deploying neural networks.