Feature selection approach for Android Malware Detection using Information Gain

Abstract

Malware was designed to damage computer systems without user knowledge. Android malware is one of the platforms that usually been attacked by this malicious software. In this paper, a feature selection approach for android malware detection using Information Gain is proposed in order the difficulty to improve the speed and accuracy of the dataset classification and detection. It is because not all features will give the same result. By having this IG, it helps to reduce the features and only the best features will be experimental in this paper. Two datasets were selected from Figshare and Malgenome. Then, these datasets are divided it into two class which are Benign and Malware. We extracted 15 features based on IG value. Then, the dataset is tested on Random Forest algorithm using WEKA tools. Our proposed feature selection approach achieved promising result with 83.4% and 71% accuracy value for Malgenome and Figshare respectively.