Prompt injection taxonomy details four attack vectors and defenses

Prompt injection, a critical vulnerability in language models, is categorized into four distinct types: direct, indirect, stored, and cross-agent. Direct injection occurs when an attacker controls user input to override system instructions, typically defended by model guardrails. Indirect injection involves hiding malicious instructions within data the model processes, such as web pages or documents, which requires runtime inspection for defense. Stored injection plants a payload in persistent memory or a RAG store for later execution, necessitating memory integrity checks. Finally, cross-agent injection happens when an infection spreads from one agent to another, defended by identity and per-hop policies. AI

IMPACT Provides a framework for understanding and defending against prompt injection attacks, crucial for securing LLM applications.

RANK_REASON Detailed taxonomy of a security vulnerability in LLMs with proposed defenses. [lever_c_demoted from research: ic=1 ai=1.0]

Read on dev.to — MCP tag →

• Cameron J.R. McClelland
• MCP
• OWASP Top 10 for LLM Applications
• Simon Willison
• Speakeasy
• SQL injection

• safety
• paper

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →