Detect knowledge base poisoning attacks before they corrupt your AI. Inspired by PoisonedRAG research (USENIX Security 2025) - 95% attack success rate.
Inspired by Document Poisoning in RAG Systems by Amin Rj
Defense layer effectiveness: Embedding Anomaly Detection (95%→20%) | Prompt Hardening (95%→85%) | Output Monitoring (95%→60%)