Using Probabilistic Model Checking to Evaluate GUI Testing Techniques

Different testing techniques are being proposed in software testing to improve systems quality and increase development productivity. However, it is difficult to determine from a given set of testing techniques, which is the most effective testing technique for a certain domain, particularly if they are random-based. We are proposing a strategy and a framework that can evaluate such testing techniques. Our framework is defined compositionally and parametric ally. This allows us to characterize different aspects of systems in an incremental way as well as test specific hypothesis about the system under test. In this paper we focus on GUI-based systems. That is, the specific internal behavior of the system is unknown but it can be approximated by probabilistic behaviors. And the empirical evaluation is based on the probabilistic model checker PRISM.