RedShell: Privacy-Preserving LLM Framework for Automated Ethical Hacking
arXiv·medium signal
Bessa et al. present RedShell, a privacy-preserving, hardware-efficient framework that fine-tunes LLMs to assist pentesters in generating offensive security code. The framework addresses the gap between dark LLMs used by attackers and ethical hacking tooling, providing automation for traditional pentesting workflows while keeping data local. For security teams, this is the first documented open framework specifically designed for LLM-assisted ethical pentesting with privacy guarantees.