A Semantic Invariant Robust Watermark for Large Language Models
This work addresses the problem of robust watermarking for LLMs, which is crucial for detecting AI-generated text in applications like content moderation, but it appears incremental as it builds on prior watermarking techniques.
The paper tackles the trade-off between attack and security robustness in watermarking algorithms for large language models by proposing a semantic invariant method that uses embeddings from another LLM to determine watermark logits, achieving robustness against synonym substitution and text paraphrasing attacks while maintaining security.
Watermark algorithms for large language models (LLMs) have achieved extremely high accuracy in detecting text generated by LLMs. Such algorithms typically involve adding extra watermark logits to the LLM's logits at each generation step. However, prior algorithms face a trade-off between attack robustness and security robustness. This is because the watermark logits for a token are determined by a certain number of preceding tokens; a small number leads to low security robustness, while a large number results in insufficient attack robustness. In this work, we propose a semantic invariant watermarking method for LLMs that provides both attack robustness and security robustness. The watermark logits in our work are determined by the semantics of all preceding tokens. Specifically, we utilize another embedding LLM to generate semantic embeddings for all preceding tokens, and then these semantic embeddings are transformed into the watermark logits through our trained watermark model. Subsequent analyses and experiments demonstrated the attack robustness of our method in semantically invariant settings: synonym substitution and text paraphrasing settings. Finally, we also show that our watermark possesses adequate security robustness. Our code and data are available at \href{https://github.com/THU-BPM/Robust_Watermark}{https://github.com/THU-BPM/Robust\_Watermark}. Additionally, our algorithm could also be accessed through MarkLLM \citep{pan2024markllm} \footnote{https://github.com/THU-BPM/MarkLLM}.