Automated Inference of Models for Black Box Systems based on Interface Descriptions

In this paper we present a method and tool to fully automatically infer data-sensitive behavioral models of black-box systems in two coordinated steps: (1) syntactical analysis of the interface descriptions, here given in terms of WSDL (Web Services Description Language), for instantiating test harnesses with adequate mappers, i.e., means to bridge between the model level and the concrete execution level, and (2) test-based exploration of the target system by means of active automata learning. The first step is realized by means of the syntactic analysis of StrawBerry, a tool designed for syntactically analyzing WSDL descriptions, and the second step by the LearnLib, a flexible active automata learning framework. The new method presented in this paper (1) overcomes the manual construction of the mapper required for the learning tool, a major practical bottleneck in practice, and (2) provides global behavioral models that comprise the data-flow of the analyzed systems. The method is illustrated in detail along a concrete shop application.