A Unified Attack Framework against Statistical Generative Watermarks for Large Language Models

Summary

This research paper presents a unified attack against statistical generative watermarks for LLMs, including KGW and its variants, GumbelSoft, DipMark, ITS and SynthID. The authors identify a score-ratio alignment between detection scores and token probability shifts, which lets a black-box attacker without secret keys or model weights produce fluent text that evades detection. Evaluated on seven watermarking schemes and three LLMs, the attack reduces detection rates to at most 0.05 while preserving text quality.

Related items

• InfoGoogle’s more personalized Gemini AI is heading to your carSame vendor · The Verge (AI)
• InfoGemini 4 Argon on Cursor: solid security score, two unique solves, but slower and more expensive | Endor LabsSame vendor · Endor Labs Blog
• MediumFrom Phishing to LLMjacking: Threat Actors Targeting Cloud Billing Accounts for LLM UsageSame vendor · Wiz Research Blog
• InfoUS bans DeepSeek AI from government use as tech tension with China continuesSame vendor · BBC Technology
• MediumFake Web3 Accelerator Uses npm Packages to Spread Malware and Hijack AI Agent TrafficSame vendor · The Hacker News