Static and dynamic malware analysis using machine learning

Abstract

Malware is a section of code written with the intention of harming a device. Attacks on the Android operating system have been on the rise of late as there are plenty of applications on the Internet that possess malware. To analyze these attacks, machine learning can be used to make the process more efficient. This paper demonstrates static and dynamic analysis of Android malware. By identifying patterns from datasets created and using a myriad of classifiers, the results have been compared to infer the most optimal method of malware analysis. Various machine learning classifier algorithms are implemented, with Random Forest and Decision Tree giving the best accuracy and F1-Score of 94% in static analysis. Support Vector Machine and Neural Network have given the highest accuracies of about 99% after implementing Principal Component Analysis in dynamic analysis.