Static Back-Stack Transition Analysis for Android

In Android, activity instances at an application’s run time are organized in a back stack, and users always interact with the top instance of the stack. Users’ interactions may trigger activity launchings, which change the state of the back stack. Therefore, back-stack transition analysis is closely related to activity transition analysis, which is fundamental for understanding Android behaviors. Existing activity transition analyses are not sensitive to back-stack-related configurations, which suppose all activities are launched with the default configuration, and thus model infeasible transitions and incorrect back-stack states when there are non-default configurations in an application. In this paper, we present a tool, BSSimulator, to perform the static back-stack transition analysis. The proposed analysis contains two types of graph constructions: the activity launching graph that represents activity transitions with extracted back-stack-related configurations, and the back-stack transition graph that simulates the changes of the back stack triggered by activity launchings and back events. The study conducted on 968 open source apps from F-droid indicates that non-default configurations are in common use. The evaluations on 35 commonly used apps validate the high-precision of the generated activity launching graphs, and indicate that the proposed back-stack transition analysis outperforms the state-of-the-art one by 29.8% in avoiding 3-length infeasible paths.