“What changed is the new CaMeL paper from DeepMind, which notably does not rely on AI models to detect attacks:”
Defeating Prompt Injections by Design
CaMeL 用控制流与数据流分离防御提示注入
针对 LLM agent 处理不可信数据时易受提示注入攻击的问题,提出 CaMeL:在 LLM 外围建立一层保护性系统,即使底层模型易受攻击也能保证安全。它显式从可信的用户查询中提取控制流和数据流,使 LLM 取回的不可信数据无法影响程序流程,并用能力(capability)机制在调用工具时执行安全策略,防止私有数据经未授权数据流外泄。在 AgentDojo 上,CaMeL 以可证明的安全性解决 77% 的任务,无防御系统为 84%。
评论者认为它是目前最可信的提示注入解法之一:不依赖 AI 去检测攻击,而是靠架构上隔离不可信数据,常被当作 agent 安全设计的参考模型。
HN 书评
推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及“CaMeL is rather interesting: Defeating Prompt Injections by Design”
“CaMeL architecture is a good reference model: it surrounds the LLM with a capability-based”
“what I consider a credible solution is the recent CaMeL paper from DeepMind:”
“Research like [1] is what I love to discover”
“One of my blog posts was the foundation of the work described in the CaMeL paper from DeepMind for example:”
看全部 13 条 →
“the CaMeL paper. Defeating Prompt Injections by Design (2025): Defeating Prompt Injections by Design”
“Edit: found it. Defeating Prompt Injections by Design”
“It reminded me of Google's CaMel work Defeating Prompt Injections by Design which introduces DSL to LLM prompt to prevent prompt injection by design”
“Prompt injection is _probably_ solvable if something like [1] ever finds a mainstream implementation and adoption”
“makes me wonder if they've implemented ideas from the DeepMind CaMeL paper: Defeating Prompt Injections by Design”
“CaMeL, from "Defeating Prompt Injections by Design", Defeating Prompt Injections by Design”
“Yes, and IIRC Google used this as part of a technique against prompt injection already [0], back when models were way more susceptible to it.”
- Defeating Prompt Injections by Design ↗kyrra · 2025-03-26 · 1 分
- Defeating Prompt Injections by Design ↗xnx · 2025-04-11 · 2 分
- CaMeL: Defeating Prompt Injections by Design ↗tomrod · 2025-04-19 · 71 分
- Defeating Prompt Injections by Design ↗theptip · 2025-04-19 · 2 分
- Defeating Prompt Injections by Design ↗redbell · 2025-04-21 · 1 分