Understanding Application Behaviours for Android Security: A Systematic Characterization

Abstract

In contrast to most existing research on Android focusing on specific security issues, there is little broad understanding of Android application run-time characteristics and their security implications. To mitigate this gap, we present the first dynamic characterization study of Android applications that targets such a broad understanding for Android security. Through lightweight method-level profiling, we have collected 33GB traces of method calls and inter-component communication (ICC) from 114 popular Android applications on Google Play and 61 communicating pairs among them that enabled an extensive empirical investigation of the run-time behaviours of Android applications. Our study revealed that (1) the Android framework was the target of 88.3% of all calls during application executions, (2) callbacks accounted for merely 3% of the total method calls, (3) 75% of ICCs did not carry any data payloads with those doing so preferring bundles over URIs, (4) 85% of sensitive data sources and sinks targeted one or two top categories of information or operations which were also most likely to constitute data leaks. We discuss the security implications of our findings to secure development and effective security defense of modern Android applications.