Designed and implemented an advanced, AI-driven honeypot system that redefines traditional deception environments. Unlike static honeypots, this system leverages a policy-gated local LLM to generate dynamic, context-aware responses — simulating realistic systems and significantly increasing attacker engagement and dwell time and analyze attackers behaviors.