KrishnaSaiReddy Patil

CivicShield introduces a novel, seven-layered defense-in-depth framework that significantly enhances the security of government-facing AI chatbots against sophisticated multi-turn adversarial attacks.

RAGShield introduces a novel, pattern-based defense system that accurately detects subtle numerical claim manipulation in government RAG systems, overcoming the inherent blind spot of embedding-based defenses.

SentinelAgent introduces a formal framework, the Intent-Preserving Delegation Protocol (IPDP), to secure federal multi-agent AI systems by verifying complex delegation chains against seven properties, ensuring robust compliance even when intent verification fails.