Security Risk Analysis of Google Play Applications
Abstract
In this work we determine the security risk posed by Google Play applications to mo-
bile device users by using statistical and machine learning techniques. A Google Play
application is characterized by its description, a category, and a set of security per-
missions required to perform its described functions. Typically mobile users install
applications based on their descriptions and don't necessarily analyze the permission
settings. Therefore, an application is de ned as unsafe or risky when some permis-
sions are accessed which provide potentially sensitive and con dential information
and not justi ed by its description. We use Stanford Topic Modeling Toolbox (TMT)
to perform topic model learning on a given dataset and then perform inferencing on
a new corpus. The testing from the training data set shows that the results obtained
from using clustering are better than the classi cation approach results. The results
also indicate that it is possible to predict with high con dence the security risk of
an application based on its permission settings and description.