Bihe Zhao

Ph.D. Student in Trustworthy Machine Learning

  PDF version

Third-year Ph.D. student at CISPA Helmholtz Center for Information Security. Research interests are trustworthy AI, data provenance for generative models, watermarking, and AI privacy.

Education

2024 – present

Ph.D., Computer Science

CISPA Helmholtz Center for Information SecuritySt. Ingbert, Germany

  • Co-supervised by Prof. Adam Dziedzic and Prof. Franziska Boenisch (SprintML lab).
  • Research on data provenance, membership and dataset inference, and watermarking for generative models.
2021 – 2024

M.Sc., Cyber Science and Technology

Beihang UniversityBeijing, China

  • Co-advised by Prof. Zhenyu Guan and Prof. Song Bian.

Experience

Jan 2022 – Apr 2023

Research Intern

SenseTime ResearchBeijing, China

  • Proposed a query-efficient model extraction attack based on public datasets that outperforms prior attacks by a large margin.
  • Revealed a non-local self-similarity signal for face forgery detection and built a unified detection framework around it (IEEE SPL).
Aug 2020 – Jan 2021

Backend Development Intern

ByteDanceBeijing, China

  • Contributed to a data annotation and management platform.
  • Developed and improved an alarm center with over 20,000 rules for detecting unusual data traffic, in Go.

Talks

Oct 2026

Agent Auditing: Identity Verification and Provenance Tracing

ML in PL 2026, Security in Machine Learning tutorialWarsaw, Poland

Slides · Video

Awards

2019

First Prize, 12th National College Student Information Security Contest

National College Student Information Security Contest (top 8%)

Face-swap detection based on video watermarking and physical unclonable functions.

First Prize, Academic Excellence Award

Beihang University (top 5%)

Excellent Student

Beihang University (top 5%)

Outstanding Student President

Beihang University (top 4%)

Publications

See the publications page for the full list with abstracts and BibTeX.