Android Malware Classification Using XGBoost On Data Image Pattern
Fauzi Mohd Darus, Noor Azurati Ahmad, Aswami Fadillah Mohd Ariffin · 2019
The popularity of Android smartphone has encouraged cybercriminals to develop malware targeting this platform. In the third quarter of 2018, the number of Android malware has increased by 40% when compared with the same quarter of year 2017. The traditional malware analysis techniques need to be improved where it must be able to detect new malware quickly and accurately. This paper proposed an Android malware classification by using 8-bit grayscale images where the images features will be extracted using GIST descriptor and later be classified using XGBoost machine learning algorithm. Classes.dex and its data section will be extracted from the Android APK files before they were converted into images. k-Nearest Neighbours and Random Forest machine learning algorithms will also be used to compare the accuracy performance. The experiments show classification using data section performed better than classification on full classes.dex files in all three machine learning algorithms.