A curated collection of research and techniques for protecting the intellectual property of Large Language Models (LLMs), including watermarking, fingerprinting, and more.
👉 View Full Documentation & Papers »
Preliminary
Model Fingerprinting
├─ Non-invasive Fingerprinting
│ ├─ Weight Space Based
│ ├─ Feature Space Based
│ │ ├─ Representation Features
│ │ ├─ Logits Features
│ │ └─ Semantic Features
│ └─ Prompt Optimization Based
└─ Invasive Fingerprinting
├─ Weight Watermarking
└─ Backdoor Watermarking
Fingerprint Transfer
Fingerprint Detection & Removal
- Comprehensive Survey: Covers both classic and cutting-edge LLM copyright protection techniques.
- Systematic Categorization: Clear taxonomy of fingerprinting and watermarking methods.
- Open Source & Community Driven: Contributions and suggestions are welcome!
We welcome contributions!
Feel free to open an issue or submit a pull request.
This project is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0). See the LICENSE file for details.
Key terms:
- ✅ Share and adapt for non-commercial purposes
- ✅ Give appropriate credit to the original authors
- ✅ Provide a link to the license
- ✅ Indicate if changes were made
- ❌ No commercial use allowed
⭐️ If you find this project helpful, please consider starring us on GitHub!
For questions or suggestions: xuzhenhua0326@zju.edu.cn