Machine Learning Based Solution for Detecting Malware Android Applications

Abstract

Smartphone usage has increased rigorously. Android is one of the most used operating systems in various smartphone worldwide. It is open-source and has chances of installing third-party applications without permission. Android is the most vulnerable operating system for a malware attack. This is a big threat to cyber security. In this paper, we make a dynamic analysis using android network traffic logs. We propose an ensemble modelled approach called XGBoost to detect malware and benign applications using the traffic. The proposed model is providing the accuracy of 92.28% and a Kappa coefficient of 0.83. Finally, some of the good set of features from android applications are outlined that helps us to label them as malware and benign. The proposed model is tested across various metrics and they are providing promising results.