Saved in:
Bibliographic Details
Main Authors: Chen, Tianyu, Lou, Jian, Wang, Wenjie
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.10030
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908857653002240
author Chen, Tianyu
Lou, Jian
Wang, Wenjie
author_facet Chen, Tianyu
Lou, Jian
Wang, Wenjie
contents As Retrieval-Augmented Generation (RAG) evolves into service-oriented platforms (Rag-as-a-Service) with shared knowledge bases, protecting the copyright of contributed data becomes essential. Existing watermarking methods in RAG focus solely on textual knowledge, leaving image knowledge unprotected. In this work, we propose AQUA, the first watermark framework for image knowledge protection in Multimodal RAG systems. AQUA embeds semantic signals into synthetic images using two complementary methods: acronym-based triggers and spatial relationship cues. These techniques ensure watermark signals survive indirect watermark propagation from image retriever to textual generator, being efficient, effective and imperceptible. Experiments across diverse models and datasets show that AQUA enables robust, stealthy, and reliable copyright tracing, filling a key gap in multimodal RAG protection.
format Preprint
id arxiv_https___arxiv_org_abs_2506_10030
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Safeguarding Multimodal Knowledge Copyright in the RAG-as-a-Service Environment
Chen, Tianyu
Lou, Jian
Wang, Wenjie
Cryptography and Security
Artificial Intelligence
As Retrieval-Augmented Generation (RAG) evolves into service-oriented platforms (Rag-as-a-Service) with shared knowledge bases, protecting the copyright of contributed data becomes essential. Existing watermarking methods in RAG focus solely on textual knowledge, leaving image knowledge unprotected. In this work, we propose AQUA, the first watermark framework for image knowledge protection in Multimodal RAG systems. AQUA embeds semantic signals into synthetic images using two complementary methods: acronym-based triggers and spatial relationship cues. These techniques ensure watermark signals survive indirect watermark propagation from image retriever to textual generator, being efficient, effective and imperceptible. Experiments across diverse models and datasets show that AQUA enables robust, stealthy, and reliable copyright tracing, filling a key gap in multimodal RAG protection.
title Safeguarding Multimodal Knowledge Copyright in the RAG-as-a-Service Environment
topic Cryptography and Security
Artificial Intelligence
url https://arxiv.org/abs/2506.10030