This paper describes an efficient approach for identifying malicious Android mobile applications through specialized static program analysis. Our solution performs offline analysis and enforces the normal properties of legitimate dataflow patterns to identify programs that violate these properties. To demonstrate the feasibility of our user-centric dependence analysis, we implement a tool to generate a data dependence graph and perform preliminary evaluation to characterize both legitimate and malicious Android apps. Our preliminary results confirm our hypothesis on the differences in user-centric data dependence behaviors between legitimate and malicious apps.