Skip to Main content Skip to Navigation

Exploiting Model Structure in CEGAR Verification Method

Abstract : Software is now a key component of majority of devices and it is responsible for their safety and reliability. By safety we mean that the system must ensure that ''bad things never happen''. This type of property can be seen as a reachability problem: to prove the property, it suffices to prove that states designated as ''bad'' cannot be reached. This is particularly important for critical systems: systems whose failure can jeopardize human life, or economic liability. We present two verifcation methods for AltaRicas models. First, a CEGAR algorithm that prunes away abstract states and therefore uses an underapproximation of the system state space is proposed. The use of our underapproximation of the abstract state space allow us to accelerate the algorithm. With our framework, we can pinpoint obvious feasible counterexamples, use reductions techniques to discard useless abstract states, minimize the cost of counterexample analysis, and guide the exploration of the abstraction towards counterexamples that are more likely to be feasible. We have implemented this framework in the model checker Mec~5, and experimental results confirmed the expected improvements. We also propose a CEGAR algorithm for a subset of the AltaRica language: we consider the situation where we want to apply CEGAR algorithm to a hierarchical transition system. We want to do this without calculating the semantics of the hierarchical system. We propose to use hierarchical abstractions where each component is abstracted independently despite the presence of priorities in the model. This has three advantages: an abstraction is represented in a succinct way, it is easy to verify if an abstract path is spurious, the abstraction reflects the logical structure of the system. Finally, we present the implementation our prunning algorithm in Mec 5. Benchmarks on a set of academic models, and on a large industrial case study illustrate the expected gain of our algorithm.
Document type :
Complete list of metadata

Cited literature [58 references]  Display  Hide  Download
Contributor : Grégoire Sutre <>
Submitted on : Monday, June 23, 2014 - 11:18:52 AM
Last modification on : Thursday, January 11, 2018 - 6:20:17 AM
Long-term archiving on: : Tuesday, September 23, 2014 - 11:00:35 AM


  • HAL Id : tel-01011163, version 1



Farès Chucri. Exploiting Model Structure in CEGAR Verification Method. Logic in Computer Science [cs.LO]. Université Sciences et Technologies - Bordeaux I, 2012. English. ⟨tel-01011163⟩



Record views


Files downloads