Comparative Analysis of Ensemble Methods for Classification of Android Malicious Applications

Abstract

Currently, Android smartphone operating systems are the most popular entity found in the market. It is open source software which allows developers to take complete benefit of the mobile operation device, but additionally increases sizable issues related to malicious applications. With the increase in Android phone users, the risk of Android malware is increasing. This paper compares the basic machine learning algorithms and different ensemble methods for classifying Android malicious applications. Various machine learning algorithms such as Random Forest, Logistic Regression, Support Vector Machine, K-Nearest Neighbor, Decision Tree and Naive Bayes and ensemble methods like Bagging, Boosting and Stacking are applied on a dataset comprising of permissions, intents, Application programming interface (API) calls and command signatures extracted from Android applications. The results revealed that the stacking ensemble techniques performed better as compared to the Bagging, Boosting and base classifiers.