arXiv cs.AI· Haoyang Liu, Yinong Xu, Zhe Lang, Zhenyu Guan, Qianyun Zhang·· 10 小时前AI 评分41
从文本到像素的转换能挡住什么?图像渲染文档中针对提示词注入的条件性防护
What Survives the Shift from Text to Pixels? Conditional Protection against Prompt Injection in Image-Rendered Documents
AI 导读
将不可信内容渲染为页面图像可降低间接提示词注入攻击成功率,但 Qwen3-VL-8B 对图像承载的答案替换操作执行率仍达 77.96%。研究对比 191 个英文问答源的文本与图像版本、覆盖六个模型,发现防护效果取决于目标操作类型、渲染设置、措辞与重复次数,且模型常能转录并识别图像中的指令却不执行。
正文 · 原文
Abstract:Multimodal language models increasingly answer questions over external documents, exposing them to indirect prompt injection: commands planted in untrusted content to hijack the response. Rendering untrusted content as page images lowers attack success on average, leaving open which injected operations survive rendering, at what utility cost, and whether failed image attacks reflect unread commands or commands read but not followed. We pair text and page-image versions of 191 English question-answering sources across six models, measuring marker appending, answer formatting, and answer replacement alongside clean Question-Answering utility. Rendering reduces append and format execution for every model, yet Qwen3-VL-8B still executes image-borne replacement at 77.96\%. Protection also depends on target type, rendering settings, wording, and repetition, and a search over eight fixed attack candidates largely closes the image deficit in appending and formatting for Qwen and Gemini, though not for GPT-5-nano. Utility ratios on original questions range from 93.31\% to 101.61\%, but Qwen falls to 61.99\% when answers combine two evidence spans. Illegibility cannot explain all failed attacks: models often transcribe and identify image-borne commands they do not execute, and trusted user authorization raises execution on unchanged pages. Inside Qwen, patching early attack-region states induces all three operations, more weakly in images. Attention blocking suggests why: a text command acts through its own positions, whereas visual encoding spreads an image command and its target across the instruction page. Rendering thus offers conditional protection, motivating evaluations that report security by model, operation, and attack budget alongside utility.
| Subjects: | Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.24231 [cs.CR] |
| (or arXiv:2611.24231v1 [cs.CR] for this version) | |
| https://doi.org/10.48550/arXiv.2611.24231 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Haoyang Liu [view email]
[v1]
Wed, 18 Nov 2026 11:08:07 UTC (790 KB)
来源:arXiv cs.AI · arxiv.org