AI-Enhanced Honeypots: Leveraging LLM for Adaptive Cybersecurity Responses
Jason Aljenova Christli, Charles Ci-Wen Lim, Yevonnael Andrew · 2024
Honeypots have long been used as decoy systems to lure and study attackers, providing valuable intelligence on emerging cybersecurity threats. However, with the increasing sophistication of cyberattacks, static honeypots have become less effective against advanced adversaries. This paper presents an innovative solution: AI-driven honeypots powered by Large Language Models (LLMs), specifically the LLaMA-3 model. Unlike traditional honeypots, these AI-enhanced systems can dynamically generate contextually appropriate, human-like responses in real-time, greatly improving their ability to deceive and engage attackers. By leveraging the LLaMA-3 model, the proposed system enhances the realism of interactions, making it significantly more difficult for attackers to identify the decoy.