Research On Activity-insulated Multi-level Testing For Android Applications

Android has become the most popular mobile system and Android applications also grow rapidly.Therefore,automated test techniques for Android applications have been widely studied.With the integration of many features,the GUI states of Android applications become more complicated.However,existing approaches to automated testing for Android applications cannot effectively explore GUI states and expose bugs.To address these problems,we propose Activity-Insulated Multi-level testing technol-ogy(AIM),which can integrate the strength of existing methods.Specifically,the work of this thesis mainly includes the following aspects:1.Propose Activity-Insulated Multi-level testing technology for Android application.With this technology,application GUI states are divided into two levels and different strategies can be adopted to explore each state level.2.Provide an exploration strategy for AIM.This strategy discovers unexplored ac-tivities based on GUI state model and then intensively exploits every discovered individual with a reinforcement learning guided random algorithm.3.Implement two test tools,AimDroid-BFS and AimDroid-DFS.4.Conduct comprehensive experiments against 50 popular closed-source commercial apps.The results demonstrate that AimDroid-BFS and AimDroid-DFS outperform both Sapienz and Monkey in activity,method and instruction coverage,respec-tively.In addition,AimDroid-BFS and AimDroid-DFS also report more crashes than the other two.