模糊测试是保障Android应用质量的一个重要手段。但Android应用的模糊测试仍然存在着诸多挑战：如何准确而快速的识别Android游戏中的可操作控件，实现有效的游戏模糊测试；如何实现可扩展的白盒模糊测试；如何解决Android应用的测试预言判定问题；如何高效的对引发错误的模糊测试用例进行精简来方便调试。这些都是我们实现高效的Android应用模糊测试系统要解决的关键问题。针对这些问题，在本项目中，我们将研究基于深度学习方法的游戏控件自动化识别技术、研究白盒模糊测试中对应用执行环境的模拟技术和集群并行加速技术、研究Android应用的差分测试技术、研究高效的层次化增量调试技术。我们前期在Android测试领域的研究基础和技术积累为本项目的成功实施提供了有力保障。而本项目的研究成果也能够进一步推动Android应用模糊测试技术的发展，从而促进移动应用质量的提高。

Fuzzing is an important technique to ensure the quality of Android application. However, the fuzzing of Android application still faces several challenges. How to recognize the rendered GUI widgets within an Android game application? How to realize scalable white-box fuzzing? How to solve the test oracle problem for Android application? How to reduce the failure-triggering input event sequence for ease of debugging? These problems must be solved when building an effective fuzzing system. In this project, we will study the automatic GUI widget recognition technique based on deep learning. We will investigate the execution environment simulation and parallel acceleration technique for white-box fuzzing. We will study differential testing technique to address the test oracle problem. And finally we will work on hierarchical delta debugging techniques for input event sequence reduction. Our previous research and technical experiences on android testing will help ensure the success of this project. The research findings of this project will improve the state of the art of Android fuzzing tools, which in turn will help improve the quality of Android applications in the market.