AI-ENHANCED HONEYPOTS FOR ADVANCED CYBER DECEPTION STRATEGIES

The use of cyber deception as a means to mitigate evolving cyber threats has become more important.Honeypots are well known for their proactive defense capabilities and have drawn considerable attention.This work researches on the integration of AI enhanced honeypots and machine learning (ML) and game theoretic models in helping honeypot deception.The proposed system utilizes Large Language Models (LLMs) to generate adaptive honeypots that can emulate real world systems very well.Further, the system can adapt to attacker's behavior in real time by incorporating Factored Interactive POMDPs.To increase in scalability and resilience, deception-in-depth architectures and container based honeypot designs were taken under consideration in recent times.We evaluate the performance of our system using simulated cyberattacks and show up to 20% improvement in the increase of attacker engagement time and up to 35% improvement in the intrusion detection rates over traditional honeypot frameworks.A scalable, adaptive honeypot framework will be developed, we will then evaluate the performance of this framework to modern cyber threats, and we will determine whether or not our framework will improve intrusion detection rates.By combining AI, ML and behavioral model this research contributes in advancing defensive deception in the form of robust and adaptive security solutions.These results help honeypot technologies to evolve in order to better adapt to complex cyber threat landscapes.