Unverified50% confidenceFactExact time
间接提示注入攻击中，攻击者只需在RAG会检索到的数据源中植入恶意指令即可操纵模型，传统WAF无法拦截
1
Sources
50%
Confidence
Long-term
Relevance
7/18/2026
First Seen
Sources
Related Claims
UnverifiedRAG架构引入了间接提示注入攻击面：攻击者只需在RAG检索的数据源中植入恶意指令即可操纵模型83% similarVerifiedIndirect prompt injection conceals attack instructions within external documents, web pages, or database records retrieved by the model, and is activated when a RAG system introduces them into the context81% similarVerified检索增强生成（RAG）是主流幻觉抑制技术，在模型生成回答前从外部知识库检索相关文档片段作为上下文注入提示词65% similarUnverified针对特定触发词的后门攻击可让模型在大多数情况下表现正常，仅在遇到特定输入时才输出预设错误信息，使常规评测基准难以捕捉64% similarUnverifiedThe essence of prompt injection attacks is exploiting the structural flaw by which LLMs cannot distinguish between instructions and data at the architectural level.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/547721API
curl https://kongchang.com/api/v1/knowledge/claims/547721MCP
get_claim(id=547721)