Proofs and Refutations for Probabilistic Refinement

  • Authors:
  • A. K. Mciver;C. C. Morgan;C. Gonzalia

  • Affiliations:
  • Dept. Computer Science, Macquarie University, Australia NSW 2109;School of Comp. Sci. and Eng., Univ. New South Wales, Australia NSW 2052;Dept. Computer Science, Macquarie University, Australia NSW 2109

  • Venue:
  • FM '08 Proceedings of the 15th international symposium on Formal Methods
  • Year:
  • 2008

Quantified Score

Hi-index 0.02

Visualization

Abstract

We consider the issue of finding and presenting counterexamples to a claim "this specis implemented by that imp", that is $\textit{spec} \sqsubseteq \textit{imp}$ (refinement), in the context of probabilisticsystems: using a geometric interpretation of the probabilistic/demonic semantic domain we are able to encode both refinement success and refinement failure as linear satisfaction problems, which can then be analysed automatically by an SMT solver. This allows the automatic discovery of certificates for counterexamples in independently and efficiently checkable form. In many cases the counterexamples can subsequently be converted into "source level" hints for the verifier.