DroidXP: A Benchmark for Supporting the Research on Mining Android Sandboxes

Due to the popularization of Android and the full range of applications (apps) targeting this platform, many security issues have emerged, attracting researchers and practitioners’ attention. As such, many techniques for addressing security Android issues have emerged, including approaches for mining sandboxes using dynamic analysis tools (i.e., automated testing tools). Undoubtedly, the resulting sandboxes’ efficiency depends on the test case generation tools used in the mining procedures. Previous research studies have compared Android test case generation tools for this specific goal. However, it is difficult to increment the research in this field because reproducing these previous empirical studies is a challenging and time-consuming task. This difficulty occurs because it is necessary to integrate test generation tools that often require different and conflicting versions of the Android platform, programming languages (e.g., Python 2 and Python 3), and software libraries. To mitigate this issue, in this paper we present DroidXP, a software infrastructure that allows researchers (and tools developers) to integrate and compare test case generation tools for mining sandboxes. We evaluated DroidXP through a reproduction study of previous research work, though considering additional test case generation tools. Our experiment suggests that DroidXP simplifies the comparison of existing tools for mining sandboxes, and revealed that Sapienz outperforms the other test case generation tools—regardless of the Monkey tool had presented the highest code coverage in our study.