An automated oracle approach to test decision-making structures

Decision-making structures are important building blocks in most of the software; however, it may be difficult to verify them because there are various input conditions and several paths causing them to behave differently. Test oracles are reliable sources of how the software must operate. The aim of the present paper is to study the applications of Artificial Neural Networks as an automated oracle to test decision-making structures. First, the decision rules were modeled by the neural network using a training dataset generated based on the software specifications and domain expert knowledge. Next, after the neural network was applied to test a subject-registration application, the proposed approach was evaluated using mutation testing. The accuracy of the resulted oracle is discussed as well.