Black-box tree test case generation through diversity

To identify defects and security risks in many real-world applications structured test cases, including test cases structured as trees are required. A simple approach is to generate random trees as test cases [random testing (RT)]; however, the RT approach is not very effective. In this work, we investigate and extend the black-box tree test case generation approaches. We introduce a novel model to produce superior test case generation based around the idea of measuring the diversity of a tree test set. This initial approach is further extended by adding a second model which describes the distribution of tree sizes. Both models are realized via a multi-objective optimization algorithm. An empirical study is performed with four real-world programs indicating that the generated tree test cases outperform test cases generated by other methods.