A Framework for Detection of Android Malware using Static Features

Abstract

Android is an open source platform that permits users to take full benefit of mobile Operating Systems, but also raises an important issue associated with malicious applications. So it is essential to make an effective technique to detect the malicious applications. In this paper, we proposed an effective framework based on the combination of static features and employ machine learning classifiers to identify malware applications. We have extracted three types of static features i.e. API calls, permissions and intents. API calls are extracted from the classes.dex and permissions and intents are extracted from AndroidManifest.xml. Then these features are used for training and testing purpose to classify the applications. The test set outcomes indicate that the combination of features performs well in comparison to the individual features and also it is found that Random Forest and K-Nearest Neighbor provide the best accuracy (i.e. 95.9%) in classifying the Android applications.