MIRAGE：针对 Web Agent 的隐蔽视觉提示注入攻击研究

Computer Science > Computer Vision and Pattern Recognition • *arXiv:2606.20717** (cs) [Submitted on 16 Jun 2026] Title:MIRAGE: Stealthy Visual Prompt Injection for Vulnerability Detection in Web Agents Authors: Xuelong Dai (https://arxiv.org/search/cs?searchtype=author&query=Dai,+X), Jianyu Ma (https://arxiv.org/search/cs?searchtype=author&query=Ma,+J), Boyang Ma (https://arxiv.org/search/cs?searchtype=author&query=Ma,+B), Biwei Yan (https://arxiv.org/search/cs?searchtype=author&query=Yan,+B), Yijun Yang (https://arxiv.org/search/cs?searchtype=author&query=Yang,+Y), Yue Zhang (https://arxiv.org/search/cs?searchtype=author&query=Zhang,+Y) View a PDF of the paper titled MIRAGE: Stealthy Visual Prompt Injection for Vulnerability Detection in Web Agents, by Xuelong Dai and 5 other authors View PDF (https://arxiv.org/pdf/2606.20717) HTML (experimental) (https://arxiv.org/html/2606.20717v1) Abstract:Multimodal Large Language Model (MLLM)-based web agents provide practical, high-precision solutions for visual browser automation; however, they inherently expand the attack surface, introducing novel vision-based vulnerabilities. Existing adversarial evaluations targeting these agents frequently rely on permissive threat models and visually conspicuous artifacts. In this paper, we investigate a constrained vulnerability detection setting: a trusted web platform where the evaluator acts solely as an unprivileged third party, such as a merchant or advertiser, controlling only a semantically legitimate, spatially constrained region, such as an ad slot, a sponsored card, or a localized widget. Operating under these realistic constraints, we propose MIRAGE, a novel visual indirect prompt injection framework for targeted next-action hijacking. Our approach leverages diffusion models to generate perceptually benign adversarial images strictly confined to the attacker-controlled boundaries permitted by the trusted service provider. To maximize attack efficacy within such a restrictive setting, we introduce a robust optimization technique combining curvature-aware adversarial diffusion guidance with sparse, dark-pixel residual perturbations. Comprehensive evaluations against prominent MLLM web agent frameworks, specifically SeeAct and OpenClaw, empirically demonstrate the potency, realism, and stealth of our proposed MIRAGE. Focus to learn more arXiv-issued DOI via DataCite (pending registration) | Submission history From: Jianyu Ma [view email (https://arxiv.org/show-email/256910cb/2606.20717)] • *[v1]** Tue, 16 Jun 2026 15:31:33 UTC (6,477 KB) Full-text links: Access Paper: View a PDF of the paper titled MIRAGE: Stealthy Visual Prompt Injection for Vulnerability Detection in Web Agents, by Xuelong Dai and 5 other authors • View PDF (https://arxiv.org/pdf/2606.20717) • HTML (experimental) (https://arxiv.org/html/2606.20717v1) • TeX Source (https://arxiv.org/src/2606.20717) view license (http://arxiv.org/licenses/nonexclusive-distrib/1.0/) Current browse context: cs.CV < prev (https://arxiv.org/prevnext?id=2606.20717&function=prev&context=cs.CV) \| next > (https://arxiv.org/prevnext?id=2606.20717&function=next&context=cs.CV) new (https://arxiv.org/list/cs.CV/new) \| recent (https://arxiv.org/list/cs.CV/recent) \| 2026-06 (https://arxiv.org/list/cs.CV/2026-06) Change to browse by: cs (https://arxiv.org/abs/2606.20717?context=cs) cs.AI (https://arxiv.org/abs/2606.20717?context=cs.AI) cs.CR (https://arxiv.org/abs/2606.20717?context=cs.CR) References & Citations • NASA ADS (https://ui.adsabs.harvard.edu/abs/arXiv:2606.20717) • Google Scholar (https://scholar.google.com/scholar_lookup?arxiv_id=2606.20717) • Semantic Scholar (https://api.semanticscholar.org/arXiv:2606.20717) export BibTeX citation Bookmark   Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer ( What is the Explorer? (https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer)) Connected Papers Toggle Connected Papers ( What is Connected Papers? (https://www.connectedpapers.com/about)) Litmaps Toggle Litmaps ( What is Litmaps? (https://www.litmaps.co/)) scite.ai Toggle scite Smart Citations ( What are Smart Citations? (https://www.scite.ai/)) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv ( What is alphaXiv? (https://alphaxiv.org/)) Links to Code Toggle CatalyzeX Code Finder for Papers ( What is CatalyzeX? (https://www.catalyzex.com/)) DagsHub Toggle DagsHub ( What is DagsHub? (https://dagshub.com/)) GotitPub Toggle Gotit.pub ( What is GotitPub? (http://gotit.pub/faq)) Huggingface Toggle Hugging Face ( What is Huggingface? (https://huggingface.co/huggingface)) ScienceCast Toggle ScienceCast ( What is ScienceCast? (https://sciencecast.org/welcome)) Demos Demos Replicate Toggle Replicate ( What is Replicate? (https://replicate.com/docs/arxiv/about)) Spaces Toggle Hugging Face Spaces ( What is Spaces? (https://huggingface.co/docs/hub/spaces)) Spaces Toggle TXYZ.AI ( What is TXYZ.AI? (https://txyz.ai/)) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower ( What are Influence Flowers? (https://influencemap.cmlab.dev/)) Core recommender toggle CORE Recommender ( What is CORE? (https://core.ac.uk/services/recommender)) • Author • Venue • Institution • Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs (https://info.arxiv.org/labs/index.html). Which authors of this paper are endorsers? (https://arxiv.org/auth/show-endorsers/2606.20717) \| Disable MathJax ( What is MathJax? (https://info.arxiv.org/help/mathjax.html))