Model-Based Testing and Some Steps towards Test-Based Modelling

Model-based testing is one of the promising technologies to increase the efficiency and effectiveness of software testing. In model-based testing, a model specifies the required behaviour of a system, and test cases are algorithmically generated from this model. Obtaining a valid model, however, is often difficult if the system is complex, contains legacy or third-party components, or if documentation is incomplete. Test-based modelling, also called automata learning, turns model-based testing around: it aims at automatically generating a model from test observations. This paper first gives an overview of formal, model-based testing in general, and of model-based testing for labelled transition system models in particular. Then the practice of model-based testing, the difficulty of obtaining models, and the role of learning are discussed. It is shown that model-based testing and learning are strongly related, and that learning can be fully expressed in the concepts of model-based testing. In particular, test coverage in model-based testing and precision of learned models turn out to be two sides of the same coin.