On the Use of Uniform Random Generation of Automata for Testing

Developing efficient and automatic testing techniques is one of the major challenges facing software validation community. In this paper, we show how a uniform random generation process of finite automata, developed in a recent work by Bassino and Nicaud, is relevant for many faces of automatic testing. The main contribution is to show how to combine two major testing approaches: model-based testing and random testing. This leads to a new testing technique successfully experimented on a realistic case study. We also illustrate how the power of random testing, applied on a Chinese Postman Problem implementation, points out an error in a well-known algorithm. Finally, we provide some statistics on model-based testing algorithms.