PI: Thomas Johansson, Lund University
Co-PI: Cristofer Englund, Halmstad University
Collaborators: Assoc. Prof. Qian Guo, Lund University, Assoc. Prof. Mahdi Fazeli, Halmstad University
Current hardware verification methods are inadequate in identifying important vulnerabilities prior to product release, as shown by famous exploits such as Spectre and Meltdown and by recent work such as GoFetch [1]. This project addresses this issue with an integrated AI-driven approach comprising two tightly coupled work packages. We will create an AI-driven discovery engine to identify microarchitectural vulnerabilities in off-the-shelf CPUs and a complementary pre-silicon verification framework to detect and mitigate them during design. The work focuses on the security of post-quantum cryptography implementations and uses a shared interface for evidence and evaluation so that post-silicon discovery informs design-time prevention. The result will be new security-related findings together with open-source tools that connect post-silicon exploit identification with pre-silicon prevention, facilitating the design of secure hardware for the future.
Project number: G6