Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Watermark Stealing in Large Language Models

About

LLM watermarking has attracted attention as a promising way to detect AI-generated content, with some works suggesting that current schemes may already be fit for deployment. In this work we dispute this claim, identifying watermark stealing (WS) as a fundamental vulnerability of these schemes. We show that querying the API of the watermarked LLM to approximately reverse-engineer a watermark enables practical spoofing attacks, as hypothesized in prior work, but also greatly boosts scrubbing attacks, which was previously unnoticed. We are the first to propose an automated WS algorithm and use it in the first comprehensive study of spoofing and scrubbing in realistic settings. We show that for under $50 an attacker can both spoof and scrub state-of-the-art schemes previously considered safe, with average success rate of over 80%. Our findings challenge common beliefs about LLM watermarking, stressing the need for more robust schemes. We make all our code and additional examples available at https://watermark-stealing.org.

Nikola Jovanovi\'c, Robin Staab, Martin Vechev• 2024

Related benchmarks

TaskDatasetResultRank
Spoofing Attack DetectionDolly CW
WCS1.333
18
Spoofing Attack DetectionHarM
WCS1.431
18
Watermark Removal AttackKGW_self 500 token (test)
ESR89
6
Watermark Attack EvaluationQwen2-7B-Instruct + KGW 2,000 tokens
RankKS Statistic0.091
5
Watermark Removal AttackKGW 500 token (test)
ESR91
3
Showing 5 of 5 rows

Other info

Follow for update