Android Mobile Malware Classification Using a Tokenization Approach

Abstract

Android is one of the most commonly targeted platforms in terms of mobile malware attacks on the part of many users worldwide. Different types of attacks and exploitations have been developed to masquerade as genuine mobile applications in order to obtain confidential information from the victim’s smartphone. Therefore, to overcome these challenges, a new mobile malware classification based on system calls and permissions using a tokenization approach is developed in this paper. The experiment was conducted in a controlled lab environment by using static and dynamic analyses to extract permissions and system calls from call logs. A total of 5560 samples from Drebin were used as training dataset, and 500 samples from Google Playstore were used as testing dataset. The new classification involving the use of a tokenization approach produced a 99.86% accuracy rate and has outperformed existing methods. This new classification can be used as guidance, and reference for other researchers with the same interests. In the future it can be used as input for the formation of a mobile malware detection model.