Automated Test Generation for Detection of Leaks in Android Applications

Android devices have limited hardware resources (e.g., memory). Excessive consumption of such resources may lead to crashes, poor responsiveness, battery drain, and negative user experience. We propose an approach for systematic automated test generation to expose resource leak defects in Android applications. We first define the notion of a neutral sequence of GUI events. Intuitively, such a sequence can be expected to have "neutral" effects on resource consumption, and repeated executions of this sequence should not exhibit resource growth. Using a state-of-the-art static control-flow model of an Android application, we demonstrate how to achieve automated generation of such sequences. We then define test generation algorithms for two important categories of neutral sequences, based on common leak patterns specific to Android. Our experimental evaluation compares this approach with a non-automated approach from prior work. The results from this evaluation strongly indicate that it is possible to achieve effective, general, and automated test generation for resource leaks in Android applications using the proposed techniques.