新研究质疑提示注入攻击对RAG系统的有效性

报道来源 [6]

1. arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Guido Zuccon · 2026-05-27 06:16

   能否抵达生成器？探究在真实 RAG 环境下提示注入攻击的存活率

   Recent generative engine optimisation (GEO) research has shown that prompt-injection attacks can push a target product to the top of an LLM's recommendation list, with the strongest attacks reporting around $80\%$ success and raising serious security concerns about RAG-based reco…

2. arXiv cs.IR (Information Retrieval) TIER_1 English(EN) · Guido Zuccon · 2026-05-27 06:16

   能否抵达生成器？探究在真实 RAG 环境下提示注入攻击的生存能力

   Recent generative engine optimisation (GEO) research has shown that prompt-injection attacks can push a target product to the top of an LLM's recommendation list, with the strongest attacks reporting around $80\%$ success and raising serious security concerns about RAG-based reco…

3. arXiv cs.AI TIER_1 English(EN) · Lei Zhao, Abhay Bhaskar, Edgar Dobriban · 2026-05-26 04:00

   LivePI：更真实地对代理进行间接提示注入基准测试

   arXiv:2605.17986v2 Announce Type: replace-cross Abstract: AI agents such as OpenClaw are increasingly deployed in local workflows with access to external tools. This creates indirect prompt-injection (IPI) risk: an agent may execute harmful instructions embedded in untrusted inpu…

4. arXiv cs.LG TIER_1 English(EN) · Zixuan Chen, Jiaxiang Chen, Li Luo, Ke Xu, Xiaoxiang Huang, Tanfeng Sun, Xinghao Jiang · 2026-05-26 04:00

   IterInject：通过反馈引导的迭代优化实现针对大语言模型代理的间接提示注入

   arXiv:2605.24659v1 Announce Type: new Abstract: LLM-based agents are increasingly deployed for complex tasks requiring planning, tool use, and interaction with external services. Their reliance on untrusted external content exposes them to indirect prompt injection (IPI), in whic…

5. Hugging Face Daily Papers TIER_1 English(EN) · 2026-05-18 07:41

   LivePI：更真实地评估代理对间接提示注入的抵抗能力

   AI agents such as OpenClaw are increasingly deployed in local workflows with access to external tools. This creates indirect prompt-injection (IPI) risk: an agent may execute harmful instructions embedded in untrusted inputs such as email, downloaded files, webpages, repositories…

6. dev.to — LLM tag TIER_1 English(EN) · Mustafa ERBAY · 2026-05-27 05:16

   AI提示注入防御：5步构建有效策略

   <p>This morning, while working on an LLM integration in my own financial analysis tool, I encountered an unintended response. While expecting a simple data query, the model spilled out a text explaining my system configuration. At first, I thought it was a bug, but upon closer in…