Our research focuses on Trustworthy Visual Intelligence:
understanding, authenticating, and protecting visual content against
manipulation, misinformation, and misuse, while advancing the theoretical
foundations of robust and secure AI-driven visual systems. We currently
pursue three directions:
-
Multimedia Forensics. Developing techniques to detect and
localize manipulated visual content, while protecting users from their
harmful effects.
-
Model Security. Identifying and mitigating threats to vision
systems, including adversarial attacks, data security risks, and
privacy leakage.
-
Visual Analysis. Advancing computer vision and machine learning
methods for reliable visual understanding and representation.