LLM-based Attack Scenarios Generator with IT Asset Management and Vulnerability Information

As businesses become more dependent on IT due to digital transformation, a variety of attackers are targeting companies, government agencies, and individuals to steal information and disrupt services. To reduce the risks these cyber threats pose, penetration testing and red teaming are important. On the other hand, these initiatives require skills and knowledge, and there is a shortage of human resources. This research aims to demonstrate the effectiveness of a system that inputs asset management data and vulnerability information into ChatGPT and searches for attack routes with a high threat level. Specifically, ChatGPT uses information used for IT asset management (OS type, version, device usage, account), vulnerability information published by CISA, and network information as input values to verify whether it is possible to output attack routes that are useful for penetration testing and red teaming. The results of the experiment confirmed that attack vectors for penetration testing and red teaming could be used to effectively uncover cybersecurity threats within an organization and perform risk assessments.