SENTINEL: generating GUI tests for sensor leaks in Android and Android wear apps

摘要

Due to the widespread use of Android devices and apps, it is important to develop tools and techniques to improve app quality and performance. Our work focuses on a problem related to hardware sensors on Android devices: the failure to disable unneeded sensors, which leads to sensor leaks and thus battery drain. We propose the Sentinel testing tool to uncover such leaks. The tool performs static analysis of app code and produces a model which maps GUI events to callback methods that affect sensor behavior. Edges in the model are labeled with symbols representing the acquiring/releasing of sensors and the opening/closing of UI windows. The model is traversed to identify paths that are likely to exhibit sensor leaks during run-time execution based on two context-free languages over the symbol alphabet. The reported paths are then used to generate test cases. The execution of each test case tracks the run-time behavior of sensors and reports observed leaks. This approach has been applied to both open-sourced and closed-sourced regular Android applications as well as watch faces for Android Wear smartwatches. Our experimental results indicate that Sentinel effectively detects sensor leaks, while focusing the testing efforts on a very small subset of possible GUI event sequences.