disarmBot — LLM Agents , disinfo

Hello everyone, nice to meet you!I am a security engineer working in Tokyo, and I am a huge fan of LLMs! I have developed a method to utilize MITRE ATT&CK using LLM and RAG, and created disarmBot as a specific PoC. disarmBot is an open-source bot that can be deployed on Discord. When users input commands, multiple LLM agents (GPT-4) automatically start up and respond. It is based on the DISARM (Disinformation Analysis and Response Measures) TTP Frameworks, which in turn are based on MITRE ATT&CK, the "theory" behind it. In other words, it is a practical measure for leveraging threat intelligence through LLMs, moving from theory to public assistance. Multiple LLM agents, each trained on different tactics, collaborate to facilitate tactical and technical dialogues based on the framework, involving attackers, defenders, users, skeptics, OSINT specialists, and solution architects. Through these dialogues, the agents delve deeper into information by discussing among themselves. disarmBot meets these conditions and provides an information environment where users can be exposed to diverse opinions. This helps users enhance their critical thinking and information digestion skills. In my Discord community, you can enjoy disarmBot in Japanese, English, and Chinese versions! I look forward to your participation.