Integration testing of distributed components based on learning parameterized I/O models

Abstract. The design of complex systems, e.g., telecom services, is usu-ally based on the integration of components (COTS). When components come from third party sources, their internal structure is usually un-known and the documentation is scant or inadequate. Our work addresses the issue of providing a sound support to com-ponent integration in the absence of formal models. We consider compo-nents as black boxes and use an incremental learning approach to infer partial models. At the same time, we are focusing on the richer models that are more expressive in the designing of complex systems. There-fore, we propose an I/O parameterized model and an algorithm to infer it from a black box component. This is combined with interoperability testing covering models of the components.