Saved in:
Bibliographic Details
Main Authors: Liu, Shuliang, Zheng, Qi, Xu, Jesse Jiaxi, Yan, Yibo, Zhang, Junyan, Geng, He, Liu, Aiwei, Jiang, Peijie, Liu, Jia, Tam, Yik-Cheung, Hu, Xuming
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.14067
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908546284650496
author Liu, Shuliang
Zheng, Qi
Xu, Jesse Jiaxi
Yan, Yibo
Zhang, Junyan
Geng, He
Liu, Aiwei
Jiang, Peijie
Liu, Jia
Tam, Yik-Cheung
Hu, Xuming
author_facet Liu, Shuliang
Zheng, Qi
Xu, Jesse Jiaxi
Yan, Yibo
Zhang, Junyan
Geng, He
Liu, Aiwei
Jiang, Peijie
Liu, Jia
Tam, Yik-Cheung
Hu, Xuming
contents Vision-language models demand watermarking solutions that protect intellectual property without compromising multimodal coherence. Existing text watermarking methods disrupt visual-textual alignment through biased token selection and static strategies, leaving semantic-critical concepts vulnerable. We propose VLA-Mark, a vision-aligned framework that embeds detectable watermarks while preserving semantic fidelity through cross-modal coordination. Our approach integrates multiscale visual-textual alignment metrics, combining localized patch affinity, global semantic coherence, and contextual attention patterns, to guide watermark injection without model retraining. An entropy-sensitive mechanism dynamically balances watermark strength and semantic preservation, prioritizing visual grounding during low-uncertainty generation phases. Experiments show 7.4% lower PPL and 26.6% higher BLEU than conventional methods, with near-perfect detection (98.8% AUC). The framework demonstrates 96.1\% attack resilience against attacks such as paraphrasing and synonym substitution, while maintaining text-visual consistency, establishing new standards for quality-preserving multimodal watermarking
format Preprint
id arxiv_https___arxiv_org_abs_2507_14067
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle VLA-Mark: A cross modal watermark for large vision-language alignment model
Liu, Shuliang
Zheng, Qi
Xu, Jesse Jiaxi
Yan, Yibo
Zhang, Junyan
Geng, He
Liu, Aiwei
Jiang, Peijie
Liu, Jia
Tam, Yik-Cheung
Hu, Xuming
Computer Vision and Pattern Recognition
Artificial Intelligence
Vision-language models demand watermarking solutions that protect intellectual property without compromising multimodal coherence. Existing text watermarking methods disrupt visual-textual alignment through biased token selection and static strategies, leaving semantic-critical concepts vulnerable. We propose VLA-Mark, a vision-aligned framework that embeds detectable watermarks while preserving semantic fidelity through cross-modal coordination. Our approach integrates multiscale visual-textual alignment metrics, combining localized patch affinity, global semantic coherence, and contextual attention patterns, to guide watermark injection without model retraining. An entropy-sensitive mechanism dynamically balances watermark strength and semantic preservation, prioritizing visual grounding during low-uncertainty generation phases. Experiments show 7.4% lower PPL and 26.6% higher BLEU than conventional methods, with near-perfect detection (98.8% AUC). The framework demonstrates 96.1\% attack resilience against attacks such as paraphrasing and synonym substitution, while maintaining text-visual consistency, establishing new standards for quality-preserving multimodal watermarking
title VLA-Mark: A cross modal watermark for large vision-language alignment model
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2507.14067