Solving the Cold Start Problem for Mobile App Permission Settings

Frederic Raber

Privacy Berlin 2016 10/2016.


Since the introduction of Android OS 6.0, users are for the first time empowered to decide over each of the permissions an app receives. Although very powerful, lay users are often overwhelmed by the number of options and the technicality of these. Prior work mainly focussed on automated suggestion of permission settings which are based on previous smartphone usage profiles. In my talk I will discuss techniques which offer a prediction without any previous smartphone usage, using input from other sources like social networks, blogs, or dedicated short privacy questionnaires as a basis for the prediction. Our results show there is a significant correlation between the user's personality according to the big five personality scores, as well as the IUIPC questionnaire, and the app permission settings he chooses. Whereas the prediction can only propose allow or disallow for most permissions, the location permission can be expressed in a more fine-grained way. For example by sharing some obfuscated location instead of the exact position, which allows to preserve the app functionality while keeping the exact location secret. I will present SafeLoc, an application which allows to obfuscate the location for an arbitrary app using different algorithms, and discuss how the optimal obfuscation algorithm and its settings can be predicted.

Weitere Links

Raber_slides.pdf (pdf, 7 MB )

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence