Welcome to the homepage of Shuo Shao!

Welcome to the homepage of Shuo Shao!

🪪About Me

I’m Shuo Shao😄, a third-year Ph.D. student with the College of Computer Science and Technology and the State Key Laboratory of Blockchain and Data Security, Zhejiang University. I am currently advised by Prof. Zhan Qin and Prof. Kui Ren. I also work closely with Dr. Yiming Li and Prof. Wenyuan Yang. I received my B.Eng. degree in Computer Science and Technology from Central South University in 2022, advised by Prof. Xiyao Liu.

My research interests focus on three main aspects of trustworthy AI: copyright, security, and privacy, including

  • AI Copyright Protection
    • Model Watermarking and Fingerprinting
    • Model Authorization and Access Control
    • Dataset Auditing
  • AI Security
    • Backdoor Attack and Defense
    • LLM Security
  • AI Privacy
    • Membership Inference and Data Extraction
    • Privacy-preserving Model Inference

🔥News

  • [2025.1]: 🎉Our paper about backdoor defense based on model reprogramming is accepted by ICLR 2025!
  • [2024.10]: 🎉Our paper about point cloud dataset ownership verification is accepted by TIFS.
  • [2024.8]: 🎉Our paper about model watermarking, entitled “Explanation as a Watermark: Towards Harmless and Multi-bit Model Ownership Verification via Watermarking Feature Attribution”, is accepted by NDSS 2025!
  • [2024.4]: 🎉Our paper “FedTracker: Furnishing Ownership Verification and Traceability for Federated Learning Model” is accepted by TDSC.
  • [2023.11]: 🎉Our paper “Watermarking in Secure Federated Learning: A Verification Framework Based on Client-Side Backdooring” is accepted by TIST.

📖Selected Publications

Image
NDSS 2025

Explanation as a Watermark: Towards Harmless and Multi-bit Model Ownership Verification via Watermarking Feature Attribution

  • We propose a novel black-box model watermarking paradigm that embeds watermarks into the explanations of the model.
  • Our paradigm is harmless and can facilitate multi-bit watermark embedding.
Image
ICLR 2025

REFINE: Inversion-Free Backdoor Defense via Model Reprogramming

  • We revisit transformation-based and backdoor-trigger-inversion-based backdoor defense and reveal their limitations.
  • We propose a novel pre-processing-based backdoor defense via model reprogramming.
Image
TDSC 2024

FedTracker: Furnishing Ownership Verification and Traceability for Federated Learning Model

  • We start the first study about the ownership traceability of FL models.
  • We propose a continual-learning-based watermark embedding method to mitigate model utility degradation.

✉️Contact

If you have any questions about my works, please feel free to contact me.

  • Email: shaoshuo_ss@zju.edu.cn; shaoshuo2000@gmail.com