Comparative Analysis of Feature Selection Methods for Detection of Android Malware

Abstract

Over the past few years, Android has been found to be the most prevalent operating system. The increase in the adoption of Android by users has led to many security issues. The amount of malware targeting Android has significantly increased. Due to the increase in the amount of malware, their detection and classification have become a major issues. Currently, the detection techniques comprise static and dynamic malware analysis. This chapter presents a comparative study of various feature selection methods through machine learning classifiers for Android malware classification. The study examines the features acquired through static malware analysis (such as command strings, permissions, intents, and API calls), and various feature selection techniques are employed to find suitable features for classifying malware to carry out the comparative analysis. The experimental results illustrate that the gain ratio feature selection approach selects relevant features for the classification of Android malware and provides an accuracy of 97.74%.