Analysis and Detection of Malware in Android Applications Using Machine Learning

Abstract

The Android Operating System, being the leading OS for mobile phone devices, is also the primary target for malicious attackers. Applications installed in Android present a way for the attackers to breach the security of the system. Therefore, it is essential to study and analyze Android applications so that malicious applications can be properly identified. Static and dynamic analyses are two major methods by which Android applications are analyzed to segregate malicious applications from the benign ones. This paper presents a study to analyze several Android applications leveraging several machine learning models. Taking different features and applying various classifiers, we show that the dynamic analysis model can hit up to 93% accuracy in detecting malware whereas the static analysis can achieve 81% of accuracy. Moreover, several trending Bangladeshi applications are analyzed as a part of this study resulting into acquisition of interesting insights.