R-Droid: Leveraging Android App Analysis with Static Slice Optimization

Abstract

Today's feature-rich smartphone apps intensively rely on access to highly sensitive (personal) data. This puts the user's privacy at risk of being violated by overly curious apps or libraries (like advertisements). Central app markets conceptually represent a first line of defense against such invasions of the user's privacy, but unfortunately we are still lacking full support for automatic analysis of apps' internal data flows and supporting analysts in statically assessing apps' behavior. In this paper we present a novel slice-optimization approach to leverage static analysis of Android applications. Building on top of precise application lifecycle models, we employ a slicing-based analysis to generate data-dependent statements for arbitrary points of interest in an application. As a result of our optimization, the produced slices are, on average, 49% smaller than standard slices, thus facilitating code understanding and result validation by security analysts. Moreover, by re-targeting strings, our approach enables automatic assessments for a larger number of use-cases than prior work. We consolidate our improvements on statically analyzing Android apps into a tool called R-Droid and conducted a large-scale data-leak analysis on a set of 22,700 Android apps from Google Play. R-Droid managed to identify a significantly larger set of potential privacy-violating information flows than previous work, including 2,157 sensitive flows of password-flagged UI widgets in 256 distinct apps.