ANDROID MALWARE DETECTION USING MACHINE LEARNING

Abstract

Smartphones have become a ubiquitous part of our lives. The popularity of Android has attracted many malicious authors to develop malware applications for numerous purposes such as data theft and other fraudulent activities that users face.. Previous research has shown that machine learning classifiers can be used to analyze permissions, which can help differentiate between malicious and benign applications on the Android platform. There are machine learning methods that use permission-based attributes to build malware detection models on Android devices. However, Android malware research is often hampered by the lack of complete and up-to-date raw malware data sets. This paper proposes to use machine learning models to classify Android malware using updated data sets. This study uses the NATICUSdroid [1] dataset and adds 2,500 Androguard apps to extract permission-based features. Five machine learning models (CatBoost, Random forest, XGBoost, Decision tree algorithm, and Support Vector Machine) are used for training. The training results show that the proposed method has an accuracy of 97,3%, 97,2%, 96,8%, 96,7%, and 96,7%, respectively. This is an auspicious result and can be applied in practice.