Summary of Protecting Copyright Of Medical Pre-trained Language Models: Training-free Backdoor Watermarking, by Cong Kong et al.
Protecting Copyright of Medical Pre-trained Language Models: Training-Free Backdoor Watermarking
by Cong Kong, Rui Xu, Weixi Chen, Jiawei Chen, Zhaoxia Yin
First submitted to arxiv on: 14 Sep 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed training-free backdoor watermarking method for Med-PLMs utilizes rare special symbols as trigger words that do not impact downstream task performance. The method embeds watermarks by replacing original embeddings with those of specific medical terms in the word embeddings layer, allowing for extraction of the watermark after fine-tuning on various medical downstream tasks. Experimental results demonstrate high fidelity and robustness against attacks while significantly reducing embedding time from 10 hours to 10 seconds. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Medical language models are super important for doctors and researchers, but they can be easily copied or stolen without permission. To keep these valuable tools safe, scientists developed a special way to hide their “fingerprint” in the model. This fingerprint is like a secret code that can only be read by someone who knows what it is. The new method uses rare words as the code and hides them in the model’s computer brain without affecting how well it works for its job. |
Keywords
* Artificial intelligence * Embedding * Fine tuning
