The buzz of Chat GPT-4 has rendered popular the concept of generative IA. The results obtained with trained LLM models in some fields are sometimes astonishing (e.g. the generation of source code).

The subject of the internship is to select an open LLM model, and to setup a complete framework within an internal local network to investigate the generation of new cybersecurity tests with IA. In a first stage, the candidate will have to analyse and benchmark LLM models. Then, he/she will have to setup a framework using the selected model inside a private network. Starting from an existing security tests library, the investigations will consist in trying to duplicate those tests with the IA and then eventually to get completely new ones.

The candidate must have good skills in software development languages such as Python, C, Rust or assembly and some basics in the security of embedded systems and processor architecture. A genuine enthusiasm for vulnerabilities research in complex systems, IA and the ability to communicate in a multidisciplinary environment will be highly appreciated.