Saran

Mobile applications are an important computing platform; securing mobile devices and their applications, is a primary concern. Unfortunately, the level of security controls exposed by mobile platforms does not typically meet DoD requirements. Effectively securing these applications requires transparently and efficiently identifying functional code blocks and adding instrumentation that satisfies DoD requirements and policies. Retrofitting security controls and policies for mobile devices and their applications is an extremely challenging problem. First, when the application employs obfuscation, identifying instrumentation points (functional code blocks) is difficult. Second, transparently retrofitting instrumentation requires sophisticated techniques that hide the side effects of injected instrumentation from the application while maintaining functionality. Third, maintaining transparency and control of instrumentation can adversely impact application performance.

Saran is a novel system that transparently and efficiently instruments Android applications. Saran supports the transparent instrumentation of the entire Android APK's DEX bytecode, which is instrumented using static binary decompilation that lifts the bytecode into an intermediate representation that facilitates program analysis and transformation. Our DEX bytecode instrumentation maintains transparency by intercepting and sanitizing reflective and other introspective, calls and provides completeness by supporting reflection. Saran enables APK instrumentation by defining a set of common events to instrument. Dynamic taint tracking via Saran instrumentation adds additional instructions to the original program to track the sensitive information sources that influence a program value. For example, suppose the introspection would like to report when location information is written to a network connection. In that case, Saran will add dynamic taint tracking instructions for all instructions on the path from a location read to a network write. This mechanism will provide unprecedented levels of precision for these types of predicated introspection directives.