Making Abstract Model Checking Strongly Preserving

  • Authors:
  • Francesco Ranzato;Francesco Tapparo

  • Affiliations:
  • -;-

  • Venue:
  • SAS '02 Proceedings of the 9th International Symposium on Static Analysis
  • Year:
  • 2002

Quantified Score

Hi-index 0.01

Visualization

Abstract

Usually, abstract model checking is not strongly preserving: it mayw ell exist a temporal specification which is not valid on the abstract model but which is instead satisfied bythe concrete model. Starting from the standard notion of bisimulation, we introduce a notion of completeness for abstract models: completeness together with a so-called partitioning propertyfor abstract models implies strong preservation for the past 碌-calculus. Within a rigorous abstract interpretation framework, we show that the least refinement of a given abstract model, for a suitable ordering on abstract models, which is complete and partitioning always exists, and it can be constructively characterized as a greatest fixpoint. This provides a systematic methodologyfor minimally refining an abstract model checking in order to get strong preservation.