Abstract
Android applications pose security and privacy risks for end-users. Early prediction of risk levels that are associated with Android applications can help Android developers is releasing less risky applications to end-users. Researchers have showed how code metrics can be used as early predictors of failure prone software components. Whether or not code metrics can be used to predict risk levels of Android applications requires systematic exploration. The goal of this paper is to aid Android application developers in assessing the risk associated with developed Android applications by identifying code metrics that can be used as predictors to predict two levels of risk for Android applications. In this exploratory research study the author has investigated if code metrics can be used to predict two levels of risk for Android applications. The author has used a dataset of 4416 Android applications that also included the applications' 21 code metrics. By applying logistic regression, the author observes two of the 21 code metrics can predict risk levels significantly. These code metrics are functional complexity and number of directories. Empirical findings from this exploratory study suggest that with the use of proper prediction techniques, code metrics might be used as predictors for Android risk scores successfully.
Included in
Information Security Commons, Software Engineering Commons, Technology and Innovation Commons
Code Metrics For Predicting Risk Levels of Android Applications
Android applications pose security and privacy risks for end-users. Early prediction of risk levels that are associated with Android applications can help Android developers is releasing less risky applications to end-users. Researchers have showed how code metrics can be used as early predictors of failure prone software components. Whether or not code metrics can be used to predict risk levels of Android applications requires systematic exploration. The goal of this paper is to aid Android application developers in assessing the risk associated with developed Android applications by identifying code metrics that can be used as predictors to predict two levels of risk for Android applications. In this exploratory research study the author has investigated if code metrics can be used to predict two levels of risk for Android applications. The author has used a dataset of 4416 Android applications that also included the applications' 21 code metrics. By applying logistic regression, the author observes two of the 21 code metrics can predict risk levels significantly. These code metrics are functional complexity and number of directories. Empirical findings from this exploratory study suggest that with the use of proper prediction techniques, code metrics might be used as predictors for Android risk scores successfully.