Android Malware Analysis Using Machine Learning Classifiers

Abstract

AbstractMalware is a big threat to mobile users nowadays. Attackers use Android applications installed on smartphones as a medium to steal our private and sensitive information including password, credentials, location, device information, etc. There is a need for a method that can detect malicious applications at a large scale quickly. For the past several years, static and dynamic processing has been used in malware identification. With the addition of machine learning methods, it further aims at reducing human effort and speeding up the analysis time. In this paper, we use machine learning approach for the identification of malicious applications. Our work applies different machine learning techniques and compares the results on the basis of their accuracies. Androguard and Strace are used for the extraction of static and dynamic features, respectively. Our analysis mainly uses app permissions, sources, sinks, presence of crypto code, reflection code, dynamic code, native codes as static features while system calls as dynamic features. We use Logistic regression, KNN, Decision Tree, SVM, Random Forest, Naïve Bayes algorithms for the classification.KeywordsMachine learningStaticDynamicMalwareAndroid