Threats to LLM Apps
The LLM threat model, prompt injection, data leakage, insecure output handling and excessive agency.
0 of 6 complete
- ▶ Up next14 min+50 XPThe LLM threat modelAssets, attackers and trust boundaries for LLM applications, mapped to the OWASP Top 10 for LLM Applications, with a…
- 15 min+50 XPPrompt injection: direct and indirectHow attackers hijack model behaviour through user input and through content your app fetches, why filtering cannot…
- Pro14 min+50 XPData leakage and system prompt exposureEight places where secrets and private data escape from LLM apps (prompts, context, retrieval, logs, caches…
- Pro14 min+50 XPInsecure output handling and excessive agencyWhy model output must be treated as untrusted input to your own systems (XSS, SQL, shell, paths, SSRF, code…
- Pro17 min+50 XPAgent and MCP security: confused deputies, tool poisoning and least privilegeThe attack surface of tool-using agents and MCP servers: confused-deputy attacks, indirect injection through tool…
- Pro14 min+50 XPModel extraction, abuse and denial of walletThreats to the service itself: model extraction and distillation theft, membership inference, resource-exhaustion and…