Guardado en:
Detalles Bibliográficos
Autores principales: Xu, Hengyuan, Xiang, Liyao, Yang, Borui, Ma, Xingjun, Chen, Siheng, Li, Baochun
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:https://arxiv.org/abs/2403.05842
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910919687143424
author Xu, Hengyuan
Xiang, Liyao
Yang, Borui
Ma, Xingjun
Chen, Siheng
Li, Baochun
author_facet Xu, Hengyuan
Xiang, Liyao
Yang, Borui
Ma, Xingjun
Chen, Siheng
Li, Baochun
contents Watermarking is a critical tool for model ownership verification. However, existing watermarking techniques are often designed for specific data modalities and downstream tasks, without considering the inherent architectural properties of the model. This lack of generality and robustness underscores the need for a more versatile watermarking approach. In this work, we investigate the properties of Transformer models and propose TokenMark, a modality-agnostic, robust watermarking system for pre-trained models, leveraging the permutation equivariance property. TokenMark embeds the watermark by fine-tuning the pre-trained model on a set of specifically permuted data samples, resulting in a watermarked model that contains two distinct sets of weights -- one for normal functionality and the other for watermark extraction, the latter triggered only by permuted inputs. Extensive experiments on state-of-the-art pre-trained models demonstrate that TokenMark significantly improves the robustness, efficiency, and universality of model watermarking, highlighting its potential as a unified watermarking solution.
format Preprint
id arxiv_https___arxiv_org_abs_2403_05842
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle TokenMark: A Modality-Agnostic Watermark for Pre-trained Transformers
Xu, Hengyuan
Xiang, Liyao
Yang, Borui
Ma, Xingjun
Chen, Siheng
Li, Baochun
Cryptography and Security
Artificial Intelligence
Watermarking is a critical tool for model ownership verification. However, existing watermarking techniques are often designed for specific data modalities and downstream tasks, without considering the inherent architectural properties of the model. This lack of generality and robustness underscores the need for a more versatile watermarking approach. In this work, we investigate the properties of Transformer models and propose TokenMark, a modality-agnostic, robust watermarking system for pre-trained models, leveraging the permutation equivariance property. TokenMark embeds the watermark by fine-tuning the pre-trained model on a set of specifically permuted data samples, resulting in a watermarked model that contains two distinct sets of weights -- one for normal functionality and the other for watermark extraction, the latter triggered only by permuted inputs. Extensive experiments on state-of-the-art pre-trained models demonstrate that TokenMark significantly improves the robustness, efficiency, and universality of model watermarking, highlighting its potential as a unified watermarking solution.
title TokenMark: A Modality-Agnostic Watermark for Pre-trained Transformers
topic Cryptography and Security
Artificial Intelligence
url https://arxiv.org/abs/2403.05842