Software System Testing as an Optimization Problem

Pawel Skruch, Marek Dlugosz, Wojciech Mitkowski, Marcin Szelest

Abstract: The paper presents an optimization-based approach to the testing of software systems. In the proposed approach, test case selection is combined with test execution, and these two activities are conducted in parallel. Executing a test case consists of exiting the system with an input and comparing the resulting output with the expected one. The test selection procedure is iterative and is driven by the optimization algorithm. The performance index to be maximized is defined as the distance between the actual and expected outputs. This formalism will allow one to find areas that have been implemented incorrectly. An empty set of test cases will give some confidence that the system has been properly implemented. The main advantage of the approach is focus on error-prone scenarios that leads to drastically reduced number of tests to be executed, and consequently to less testing effort. An example is included to verify the approach and to present its applicability for industrial projects.