Fuzzing Hardware: Faith or Reality? : Invited Paper
Weimin Fu, Orlando Arias, Yier Jin, Xiaolong Guo · 2021
Compared to software defects which can be patched in the field, hardware defects are permanent. As hardware iterations accelerate, a leftward shift in hardware testing is necessary. Among all existing techniques, formal methods (both automated and deductive) are the most effective solutions in detecting vulnerabilities in hardware. However, most of the existing formal methods are not scalable to large-scale designs due to the lack of practical automated tools. Very recently, hardware fuzzing solutions are proposed which treat the executable simulation code directly as software and test it with a Fuzz tool such as AFL or Symbolic Execution Engine such as KLEE. In this paper, we survey existing hardware fuzzing studies and discuss whether it is a valuable research direction to pursue. We also review these approaches by identifying potential challenges and gaps, based on which we present visions that can be performed to eliminate these challenges.