LLM application security¶
All topics · Alphabetical index
Concepts¶
- Direct prompt injection; indirect prompt injection; tool-output injection
- Instruction-data separation; trust boundaries; untrusted documents
- Retrieval poisoning; memory poisoning
- Secret leakage; data exfiltration; cross-tenant leakage
- Excessive agency; privilege escalation; confused-deputy attacks
- Malicious URLs; SSRF; generated command injection
- Sandbox isolation; network egress controls; output validation
- Approval bypass; tool supply chain; connector permissions
- Jailbreaks; adversarial evaluation; policy enforcement
Further reading¶
Reference links for this topic.