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...

Full description

Saved in:
Bibliographic Details
Published inIEEE access Vol. 7; pp. 110781 - 110793
Main Authors Liu, Ao, Guo, Chenkai, Wang, Weijing, Qiu, Yu, Xu, Jing
Format Journal Article
LanguageEnglish
Published Piscataway IEEE 2019
The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
Subjects
activity transition
Analytical models
Android
back stack
Configurations
Generators
Graphical user interfaces
Object oriented modeling
static analysis
Task analysis
Online AccessGet full text

Cover

More Information
Summary: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.
ISSN:2169-3536
2169-3536
DOI:10.1109/ACCESS.2019.2934528