Using Learning Techniques to Generate System Models for Online Testing

Abstract

Today’s software systems are mostly modular and have to be changeable. However, the testing of such systems becomes difficult, especially when changes are applied after deployment. One way to passively test such a system is to check whether the observed traces are accepted by a system model. In this paper, we present a method to generate a model of the System Under Test from its test cases. We adapt Angluin’s algorithm for learning finite automata to the special case of learning from traces ob- tained from test cases and provide the promising results of our experiment.