Large Language Models for Hardware Security (Invited, Short Paper)
Hammond Pearce, Benjamin Tan
Abstract
Secure digital hardware is the foundation of secure systems. However, achieving hardware security requires a lot of disparate expertise, ranging from knowledge of tools, awareness of myriad threats, and a fundamental understanding of how digital hardware is used in a given application. Such expertise is rare, so mistakes are made. We believe that recent advancements in AI, specifically Large Language Models (LLMs), provide a potential pathway for alleviating the burden on digital hardware designers. In this paper, we outline some of the challenges faced by cybersecurity experts in the hardware space, our vision for how LLMs can revolutionize hardware security, briefly present current progress and recent works in this research direction, and finally reflect on open challenges that remain for the community.