Integration of Multi-modal Features for Android Malware Detection Using Linear SVM

摘要

In light of the rapid growth of malware threats towards the Android platform, there is a pressing need to develop effective solutions. In this paper we explorate the potential of multi-modal features to enhance the detection accuracy while keep the false alarms low. Examined features include the permissions, Application Programming Interface (API) calls, and meta features such as the category information and Application Package (APK) descriptions. These multi-modal features are coded in a way to facilitate efficient learning and testing with the particular classifiers known as the linear support vector machine (SVM). Experiments show that our proposed method can obtain an accuracy more than 94%, over performing the conventional methods by a large margin. By employing high-performance learning tools, the training and testing can be done in a very time-efficient fashion for large scale and high-dimensional data.