Can Watermarks Survive Translation? On the Cross-lingual Consistency of Text Watermark for Large Language Models

Zhiwei He, Binglin Zhou, Hongkun Hao, Aiwei Liu, Xing Wang +3 more
2/21/2024
cs.CLcs.AI

Official code: zwhe99/X-SIR (49 stars, last commit 2024-06-04).

Abstract

Text watermarking technology aims to tag and identify content produced by large language models (LLMs) to prevent misuse. In this study, we introduce the concept of cross-lingual consistency in text watermarking, which assesses the ability of text watermarks to maintain their effectiveness after being translated into other languages. Preliminary empirical results from two LLMs and three watermarking methods reveal that current text watermarking technologies lack consistency when texts are translated into various languages. Based on this observation, we propose a Cross-lingual Watermark Removal Attack (CWRA) to bypass watermarking by first obtaining a response from an LLM in a pivot language, which is then translated into the target language. CWRA can effectively remove watermarks, decreasing the AUCs to a random-guessing level without performance loss. Furthermore, we analyze two key factors that contribute to the cross-lingual consistency in text watermarking and propose X-SIR as a defense method against CWRA. Code: https://github.com/zwhe99/X-SIR.

View on arXivView PDF

Code Implementations(1)

zwhe99/X-SIROfficial100%

[ACL 2024] Can Watermarks Survive Translation? On the Cross-lingual Consistency of Text Watermark for Large Language Models

496Shell, PythonFeb 23, 20242 years ago
baichuanbaichuan-7bbaichuan2baichuan2-7blarge-language-models+9 more

Similar Papers

Evaluating the Robustness and Accuracy of Text Watermarking Under Real-World Cross-Lingual Manipulations

92% match1 repo12025

Is Multilingual LLM Watermarking Truly Multilingual? A Simple Back-Translation Solution

89% match2025

Cross-Lingual Summarization as a Black-Box Watermark Removal Attack

89% match2025

Signature vs. Substance: Evaluating the Balance of Adversarial Resistance and Linguistic Quality in Watermarking Large Language Models

89% match2025

On the Reliability of Watermarks for Large Language Models

86% match1 repo6792023

On the Reliability of Watermarks for Large Language Models

86% match1 repo6792026

Cite this paper

@article{he2024watermarks,
  title  = {Can Watermarks Survive Translation? On the Cross-lingual Consistency of Text Watermark for Large Language Models},
  author = {Zhiwei He and Binglin Zhou and Hongkun Hao and Aiwei Liu and Xing Wang and Zhaopeng Tu and Zhuosheng Zhang and Rui Wang},
  year   = {2024},
  eprint = {2402.14007},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2402.14007v2}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2402.14007.svg)](https://codeofpaper.com/paper/2402.14007)

Discussion