LeakDAF: An Automated Tool for Detecting Leaked Activities and Fragments of Android Applications

Memory leak, one of the most common problems threatening android apps, might drain the limited memory of mobile devices, cause unexpected delays, no-responses or even crashes to apps. Activity/Fragment Leak is one of the most common and serious causes of memory leaks and has a vast influence on the Android app market. Existing work to identify leaked activities/fragments either depend highly on the experience of developers, or require app's source code and manual interactions. In this paper, we propose an automatic tool named LeakDAF for detecting leaked activities/fragments automatically without manual intervention. LeakDAF makes use of UI testing technique to execute automatically the app under test, and applies memory analysis technique to inspect dumped heap files to identify leaked activities and fragments based on Android's mechanisms for managing them. To evaluate the effectiveness of LeakDAF, we successfully applied it to 35 open source and 64 commercial apps, and we detected at least one leaked activity or fragment for 10 open source apps and 35 commercial apps.