ASAP: A Dynamic & Proactive Approach for Android Security Analysis and Privacy

We present a solution that applies state of the art Deep Learning methods in order to classify applications regarding their privacy footprint. For this purpose, we consider the user privacy, as stated in their public manifest, which presents reliance to popular attacks involving droppers, or application specifically design to abuse user data. The solution is carefully detailed, including the most important design choices, and validated against two standard datasets. The results demonstrate the effectiveness of our approach, yielding high F1-score and MCC metrics when compared against reference solutions. We also discuss further work and potential contributions that could be derived to enhance awareness around the privacy impact of mobile applications.