Publicly-Detectable Watermarking for Language Models
IACR Communications in Cryptology · 13.01.2025 · 1 (4)
DOI: 10.62056/ahmpdkp10 ↗Kurz gefasst
Fachartikel in der Zeitschrift IACR Communications in Cryptology, online erschienen am 13.01.2025. Verfasst von Jaiden Fairoze und 5 weiteren Autor:innen. Die Arbeit wurde bisher 15-mal zitiert (Crossref) und nennt 47 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
We present a publicly-detectable watermarking scheme for LMs: the detection algorithm contains no secret information, and it is executable by anyone. We embed a publicly-verifiable cryptographic signature into LM output using rejection sampling and prove that this produces unforgeable and distortion-free (i.e., undetectable without access to the public key) text output. We make use of error-correction to overcome periods of low entropy, a barrier for all prior watermarking schemes. We implement our scheme and find that our formal claims are met in practice.
Zitieren
APA
Fairoze, J., Garg, S., Jha, S., Mahloujifar, S., Mahmoody, M., & Wang, M. (2025). Publicly-Detectable Watermarking for Language Models. IACR Communications in Cryptology, 1(4). https://doi.org/10.62056/ahmpdkp10
Vancouver
Fairoze J, Garg S, Jha S, Mahloujifar S, Mahmoody M, Wang M. Publicly-Detectable Watermarking for Language Models. IACR Communications in Cryptology. 2025;1(4). doi:10.62056/ahmpdkp10
BibTeX
@article{Fairoze2025,
author = {Fairoze, Jaiden and Garg, Sanjam and Jha, Somesh and Mahloujifar, Saeed and Mahmoody, Mohammad and Wang, Mingyuan},
title = {Publicly-Detectable Watermarking for Language Models},
journal = {IACR Communications in Cryptology},
year = {2025},
volume = {1},
number = {4},
doi = {10.62056/ahmpdkp10}
}
EUVIMED