Two effective methods to detect mobile malware

Malware in Android has enjoyed a prevalence as Android has taken up a primary market in the mobile devices. In this paper, Adaptive Regularization Of Weights (AROW) and Support Vector Machine (SVM) are used to identify malware with both the static and dynamic analysis. Permissions, control flow graphs and system calls are taken as the application's classification features. Single features, the combination of permissions and control flow graphs, and the combination of the three features are all analyzed thoroughly. Compared with the previous work, AROW and SVM with the combination of the features could increase the TPR and decrease the FPR. Furthermore, the results of AROW and SVM are profoundly evaluated in this paper. As regard to the single features, SVM could perform better with the feature of permissions or system calls, while AROW provides a better result with control flow graphs. For the combination of the features, AROW gives a better result than SVM.