Weighted ensemble based extra tree for permission analysis for android applications classification

Abstract

Selecting optimal features for classification task is one of the essential problems in machine learning field. Feature Selection is one of the most extensively studied methods for dimensionality reduction. The feature selection method preserves a subset of the existing features and discards the rest during the (supervised or unsupervised) learning process. However, representing features plays important role in obtaining the highly discriminant features that contribute in enhancing the classifier performance. Therefore, the aim of this paper is to propose a framework based on ensemble extra tree algorithm to assign weight to features that have high influence in classifying android apps to malware or non-malware with lower computational cost overhead. The presented framework is evaluated by using different machine learning classifiers to examine the permissions features of two datasets in terms of their representation as binary vector or weight vector in enhancing the classification performance. The experimental results show that the presented model based on features weighting approach improved the classification performance.