Automated model-based android GUI testing using multi-level GUI comparison criteria

Abstract

Due to the widespread propagation of smart mobile devices, the marketplace of mobile applications (apps) has grown exponentially, and the mobile apps also could be prone to be faulty as the complexity of mobile apps has increased. An automated graphical user interface testing (GUI testing) is one of the most widely used techniques that exercises the behaviors of an application under test (AUT) by executing events on GUIs and checking if the software behaves correctly. In particular, because Android platforms have led the market share of mobile OS platforms, a lot of researches on automated Android GUI testing techniques have been performed. Among various techniques, we focus on model-based Android GUI testing that utilizes a GUI model for effective test generation and systematic debugging support. In order to generate effective test inputs in model-based approaches, accurate modeling of AUT’s behaviors is the most important factor because test input generation is based on the GUI model. However, most modern Android apps contain a number of dynamically constructed GUIs that make modeling behaviors more challenging. To address this modeling problem, we propose a Multi-level GUI Comparison Criteria (GUICC) that provides the selection of multiple abstraction levels in GUI model generation. We also have implemented a testing framework that automatically performs model-based GUI testing for Android apps with our proposed GUICC. Our experi-mental results show that we can achieve significantly higher effectiveness than existing activity-based GUI model generation, and we also find that the proposed technique can successfully alleviate the inherent state explosion problems for modeling the behaviors of real-world Android apps.