LLM Agents for Cyber Defense — 42p — Franklin Parker, Simon Paulson, Franklin Parker, Shaik Fareed, Lovisa Ivarsson, Mahesh Kamepalli

In the evolving field of cybersecurity, new threats continually emerge, challenging existing defenses. The emergence of large language models (LLMs) for executing cyberattacks allows for quicker and more complicated breaches while lowering the barrier of entry for attackers. For our poster we plan to present on how we explore the integration of LLMs with honeypot systems to bolster cybersecurity defenses. We propose enhancements to current honeypots by leveraging LLMs for deployment, management, and monitoring. Additionally, we demonstrate the use of LLMs to interpret logs into easy to understand language. Our findings indicate the incorporation of LLMs into honeypot systems enhance their realism, reduce the need for manual configurations, and adapt to evolving attacks.

Dakota State University
Mark Spanier