Modeling App Behavior from Multiple Artifacts

Abstract

Understanding the behavior of a mobile application is typically obtained from static or dynamic artifacts including an app's market page. However, each of these artifacts have limitations which make the resulting models of behavior incomplete. In response, this work asks can a technique which combines a larger, more diverse set of artifacts extracted from an app's executable, runtime behavior and market page construct more accurate models of app behavior? And if so, which artifacts and features are most valuable when doing so?