Guardado en:
Detalles Bibliográficos
Autores principales: Huang, Baizhou, Pu, Xiao, Wan, Xiaojun
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:https://arxiv.org/abs/2411.01222
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866917829784109056
author Huang, Baizhou
Pu, Xiao
Wan, Xiaojun
author_facet Huang, Baizhou
Pu, Xiao
Wan, Xiaojun
contents Watermarking has emerged as a prominent technique for LLM-generated content detection by embedding imperceptible patterns. Despite supreme performance, its robustness against adversarial attacks remains underexplored. Previous work typically considers a grey-box attack setting, where the specific type of watermark is already known. Some even necessitates knowledge about hyperparameters of the watermarking method. Such prerequisites are unattainable in real-world scenarios. Targeting at a more realistic black-box threat model with fewer assumptions, we here propose $B^4$, a black-box scrubbing attack on watermarks. Specifically, we formulate the watermark scrubbing attack as a constrained optimization problem by capturing its objectives with two distributions, a Watermark Distribution and a Fidelity Distribution. This optimization problem can be approximately solved using two proxy distributions. Experimental results across 12 different settings demonstrate the superior performance of $B^4$ compared with other baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2411_01222
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle $B^4$: A Black-Box Scrubbing Attack on LLM Watermarks
Huang, Baizhou
Pu, Xiao
Wan, Xiaojun
Computation and Language
Watermarking has emerged as a prominent technique for LLM-generated content detection by embedding imperceptible patterns. Despite supreme performance, its robustness against adversarial attacks remains underexplored. Previous work typically considers a grey-box attack setting, where the specific type of watermark is already known. Some even necessitates knowledge about hyperparameters of the watermarking method. Such prerequisites are unattainable in real-world scenarios. Targeting at a more realistic black-box threat model with fewer assumptions, we here propose $B^4$, a black-box scrubbing attack on watermarks. Specifically, we formulate the watermark scrubbing attack as a constrained optimization problem by capturing its objectives with two distributions, a Watermark Distribution and a Fidelity Distribution. This optimization problem can be approximately solved using two proxy distributions. Experimental results across 12 different settings demonstrate the superior performance of $B^4$ compared with other baselines.
title $B^4$: A Black-Box Scrubbing Attack on LLM Watermarks
topic Computation and Language
url https://arxiv.org/abs/2411.01222