Automated generation of attack trees using LLMs and prompt engineering

Dr. Buttyán Levente

Attack trees are used to describe and analyze possible ways to attack a system. Generating them can be a tedious work if done manually, so we would like to automate the process. We believe that LLMs could be used to generate such attack trees, but not sure how. The task of the student is to get familar with the concept of attack trees, read the literature on automted generation of them, get some hands-on experience with using LLM-based bots to understand how to interact with them, and then propose possible ways to teach and then use a bot to help in generating attack trees for a system and an attack target given as input. This project is related to an ongoing research project in the CrySyS Lab, and the student will have the opportunity to work together with other researchers in the lab on this project.