Android Static Malware Detection using tree-based machine learning approaches

Abstract

In a world where nearly everything is connected directly to mobile phones, privacy and security are being prioritized for the safe use of smartphones. Most of the smartphones used around the world run on android operating systems nowadays. Also, with thousands of applications developed daily, Malware detection needs to be automated as human inspection for each application became impossible. Malware detection for android applications consists of two types: Static and dynamic detection. This paper is concerned with static malware detection through applications' requested permissions of use. The used dataset consists of 400K applications at which the application's permissions were extracted to numeric features. In our approach, we had 15 classes in our data which motivated us to use tree-based machine learning algorithms. Our proposed models were Random-Forest and Xgboost which got 93.39% and 93.54% respectively. We were able to achieve higher accuracies than existing models with less computational power and in a more efficient manner.