Android Malware Detection and Family Classification

Abstract

Android has become a predominant mobile operating system lately. Google with the help of Open Handset Alliance setting out to create open standards for smart phones have prompted a gigantic development in the digital world. Given the growth and development of smartphone devices and their related application stores, Malware detection is a developing issue. Volume of new applications is excessively enormous to physically analyze every application for malicious activity. Keeping this in view this research presents a method to detect android malware and further classify it to four malware categories and thirty nine malware families. The classification model has been built around reduction of redundant features and employing three machine learning algorithms (Random Forest, KNN and SVM algorithms) in binary classification and Random forest algorithm for category and family classification. The proposed methodology performs reasonably well for most of the classes achieving around an accuracy of 95% on binary classification. Proposed method provides the accuracy of 84% on malware category classification and accuracy of 66% for Malware family classification.