Statistical Model Checking: Black or White?

Pranav Ashok, Przemyslaw Daca, Jan Kretínský, Maximilian Weininger

Abstract: One of the advantages of statistical model checking (SMC) is its applicability to black-box systems. In this paper, we discuss the advantages gained when SMC is applied to white-box systems, utilizing the knowledge of their internals. We focus on the setting of unbounded-horizon properties such as reachability or LTL. We compare our approach to other statistical and numerical techniques both conceptually as instantiations of the same framework, and experimentally. It not only clearly preserves scalability advantages of black-box SMC compared to classical model checking (while providing high level of guarantees), but it also scales yet better than either of the two for a wide class of models.