This topic focuses on enhancing ARACNE, an agent-based cybersecurity framework integrating Large Language Models (LLMs) to simulate intelligent attackers and improve interaction with users. It aims to deliver clearer summaries of attack patterns, provide benchmarks for adversarial agents, and challenges to evaluate attacker adaptability and model robustness in cybersecurity environments.
LLM-based Honeypots
This topic is an extension of current Large Language Model (LLM)-based honeypots developed in the Stratosphere Lab by incorporating a wider range of services, integrating newer and more advanced language models, including locally hosted ones, and enhancing their deception capabilities. The goal is to create more realistic and adaptive honeypot systems that can effectively engage, mislead, and analyze malicious actors, while leveraging AI and data science to improve threat intelligence and system resilience.