A Configurable CEGAR Framework with Interpolation-Based Refinements

Correctness of software components in a distributed system is a key issue to ensure overall reliability. Formal verification techniques such as model checking can show design flaws at early stages of development. Abstraction is a key technique for reducing complexity by hiding information, which is not relevant for verification. Counterexample-Guided Abstraction Refinement (CEGAR) is a verification algorithm that starts from a coarse abstraction and refines it iteratively until the proper precision is obtained. Many abstraction types and refinement strategies exist for systems with different characteristics. In this paper we show how these algorithms can be combined into a configurable CEGAR framework. In our framework we also present a new CEGAR configuration based on a combination of abstractions, being able to perform better for certain models. We demonstrate the use of the framework by comparing several configurations of the algorithms on various problems, identifying their advantages and shortcomings.