Improving Search-Based Android Test Generation Using Surrogate Models.

The increasing popularity of mobile apps implies a need for automated test generation techniques for Android apps. Unlike other domains where automated test generation has been applied successfully, such as unit test generation, test execution for Android apps is computationally expensive: Tests are executed in an emulator, the app under test needs to be restarted after every test execution, and even individual actions within a test may take in the range of seconds to execute. This is inhibitive for approaches that rely on frequent execution of tests, such as search-based testing, which requires test executions to calculate fitness values. A common approach in evolutionary search is to use surrogate models as a means to reduce the costs of fitness calculation. In this paper, we introduce an approach to integrate surrogate models for testing Android apps: The surrogate model is an abstraction of the state-based behaviour of the graphical user interface, and can predict traces for already explored behaviour, thus avoiding costly test executions. We integrate this surrogate model in the search-based test generator MATE and perform an empirical study on a set of 10 Android apps. Results indicate that both the number of evaluated test cases and the resulting coverage can be increased significantly.