arXivDaily arXiv每日学术速递 周一至周五更新
​通知:联邦节假日调休:哥伦布日 / 原住民日(Columbus Day / Indigenous Peoples' Day),今日(13日)无更新。
arXiv 2505.13957cs.CRcs.CL

超越文本:揭示多模态检索增强生成中的隐私漏洞

Beyond Text: Unveiling Privacy Vulnerabilities in Multi-modal Retrieval-Augmented Generation

Michigan State University(密歇根州立大学);Jilin University(吉林大学)

Jiankun Zhang, Shenglai Zeng, Jie Ren, Tianqi Zheng, Hui Liu, Xianfeng Tang, Hui Liu, Yi Chang

AI总结:

该研究首次系统分析多模态检索增强生成(MRAG)在视觉-语言、语音-语言模态下的隐私漏洞,提出黑盒场景下的组合式结构化提示攻击方法,证实攻击者可通过操纵查询提取隐私信息,凸显研发隐私保护MRAG技术的紧迫性。

摘要:

多模态检索增强生成(MRAG)系统通过整合外部多模态数据库来增强大语言多模态模型(LMMs),但也带来了尚未被探究的隐私漏洞。尽管基于文本的RAG隐私风险已有研究,多模态数据却带来了独特挑战。我们首次对视觉-语言和语音-语言模态下的MRAG隐私漏洞展开系统性分析。我们在黑盒场景下采用一种新颖的组合式结构化提示攻击,证明了攻击者可通过操纵查询提取隐私信息。实验表明,LMMs既能直接生成与检索内容相似的输出,也能生成间接泄露敏感信息的描述,凸显了研发稳健的隐私保护MRAG技术的迫切性。

英文摘要:

Multimodal Retrieval-Augmented Generation (MRAG) systems enhance LMMs by integrating external multimodal databases, but introduce unexplored privacy vulnerabilities. While text-based RAG privacy risks have been studied, multimodal data presents unique challenges. We provide the first systematic analysis of MRAG privacy vulnerabilities across vision-language and speech-language modalities. Using a novel compositional structured prompt attack in a black-box setting, we demonstrate how attackers can extract private information by manipulating queries. Our experiments reveal that LMMs can both directly generate outputs resembling retrieved content and produce descriptions that indirectly expose sensitive information, highlighting the urgent need for robust privacy-preserving MRAG techniques.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑