Abstract: This paper proposes a semantic-based sensitive behavior analysis method for Android. With sample statistics results, the method firstly adopts a simple-Dalvik intermediate language (SDIL) as the intermediate language for text analysis, thus giving a symbolic semantics description for instructions. Then the method uses SDIL to detect sensitive calls from the samples and traces the call paths according to the control dependence. Then based on control-flow analysis, the method adopts constraint solving to obtain path conditions. At last, the method finds the background behaviors with trigger conditions, thus the whole process of background behavior execution will be showed as well. This method can release the path explosion problem in the process of symbolic execution. With experiment under our platform, it proves that the method can analyze the background behaviors of mobile application efficiently, and find the unknown mobile malicious applications which can not be found by traditional feature detection methods in time.