Our dataset is extracted from the top 50 free apps of 34
different categories (26 and 8 in Applications and Games
groups, respectively) of Google Play [26] in March 2012. The
dataset includes Android Application Package (APK) files,
decompiled to an .xml file via apktool [27]. APK’s Manifest
.xml file describes an app’s name, version, access rights, and
permission labels.
For the time being, Android includes 124 permission labels,
formulated as 124 dimensional binary vectors one per app. In
the vectors, a “1” represents a requested permission and a “0”
shows that the permission was not requested. To gain an insight
into nuance of the dataset, a statistical analysis of requested
permissions is performed as in Table II, which quantifies
distinct requested permissions of each category and the highest
number of permissions requested by apps in each category.
Amidst the categories, Games and Applications’ Widgets
requested the minimal (17) and maximal (75) distinctive
permissions. Among the apps, Productivity categories NQ
Mobile Security and Antivirus app solicits 40 permissions,
which alongside Productivity 71 unique permission requests
indicate a high diversity of permission requests.