Uncovering the Hidden Threat of Text Watermarking from Users with Cross-Lingual Knowledge

Mansour Al Ghanim Jiaqi Xue Rochana Prih Hastuti Mengxin Zheng Yan Solihin Qian Lou · 2025

Read Uncovering the Hidden Threat of Text Watermarking from Users with Cross-Lingual Knowledge by Mansour Al Ghanim Jiaqi Xue Rochana Prih Hastuti Mengxin Zheng Yan Solihin Qian Lou (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-16699.txt.

Description

Mansour Al Ghanim and co-authors tested how well hidden marks in machine-written text survive attacks that use more than one language, in a 2025 study. Such marks help prove that text came from a model. Most past tests covered English alone, so tricks based on translation had barely been checked.

The team ran four marking methods across four rich languages. They checked how clean the marked text stayed and whether the marks could still be found after translation attacks. Their results show where current methods hold up and where they fail. From those results, the authors draw lessons for building marks that stay sound when text moves between languages.

Catalog record

CREATORMansour Al Ghanim Jiaqi Xue Rochana Prih Hastuti Mengxin Zheng Yan Solihin Qian Lou
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE50 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-16699

Browse the full library · About the Tower of Babel