PlusApps: Towards a Privacy Risk Analysis for Android Plus Applications

Abstract

The Android platform leads the mobile operating system marketplace and subsequently has drawn the interest of malware authors and researchers. The significant number of proposed malware detection techniques, classification models and practical reverse engineering solutions are insufficient and there is a lack of perfection. Also, the number of Android apps has increased significantly in recent years, as has the number of apps revealing confidential data. It is essential to investigate the applications and make sure that none of them are leaking privacy data, and consequently a privacy leak analysis approach is needed. Therefore, this paper investigates plus apps behavior and data leakages with a machine-learning algorithm to determine the best features for differentiating plus apps from original apps. The result of the analysis discloses that the SVM classifier presents the greatest accuracy. Further investigation demonstrates that the classifier with the ranking algorithm that uses correlation coefficient (CorEvel) and information gain (InfGain) methods offers more exceptional precision than the other correlation algorithms. The result of this experiment proves that the ranking algorithm is able to decrease the dimension of features and produce an accuracy of 96.60%.