Efficient resolution of security-sensitive values in Android using abstract interpretation

Abstract

In this thesis I present a design for an efficient and sound abstract interpretation-based Value Analysis which calculates field values of security-relevant Android API class instances. The analysis is an important component of DroidSafe, an Android malware detection system designed to prove important properties of sensitive program behaviors before the programs appear in an application marketplace. The resolved program values provide important context for other DroidSafe analyses and the generated application summary, improving their precision. This in turn helps a trusted analyst avoid false positives and determine whether a particular application is malicious in a shorter amount of time.