Binary and multi-class classification of Android applications using static features

Android has dominated the smart phone market in recent years. As a result, there is a massive increase in Android applications. Due to the increasing number of applications and users' dependence on these, Android has become a prime target for attackers. Hence, there is a need for new malware detection methods. Machine learning algorithms are being used for this purpose. This paper proposes a framework which is capable of performing binary and multi-classification of Android applications. Machine learning algorithms are used on a self-created dataset to classify Android applications into malicious and benign. Further, the malicious applications are classified into their families using the same machine learning algorithms. It is concluded that the accuracy of classification of malicious applications into the families gives very less accuracy (86.70% achieved by Random Forest) as compared to the binary classification accuracy (96.50% achieved by Random Forest).