Dynamic test profiles in adaptive random testing: A case study

Random testing (RT) is a basic software testing method. When used to detect software failures, RT usually generates random test cases according to a uniform distribution. Adaptive random testing (ART) is an innovative approach to enhancing the failure-detection capability of RT. Most ART algorithms are composed of two independent processes, namely the candidate generation process and the test case identification process. In these ART algorithms, some program inputs are first randomly generated as the test case candidates; then test cases are identified from these candidates in order to ensure an even spread of test cases across the input domain. Most previous studies on ART focused on the enhancement of the test case identification process, while using the uniform distribution in the candidate generation process. A recent study has shown that using a dynamic test profile in the candidate generation process can also improve the failure-detection capability of ART. In this paper, we develop various test profiles and integrate them with the test case identification process of a particular ART algorithm, namely fixed size-candidate-set ART. It is observed that all these test profiles can significantly improve the failure-detection capability of ART.