Improved Malware Identification from Android Apps using Image Features

The type and volume of malware is increasing rapidly. In this work we present an improved approach for malware classification from the android applications (or apps) using image features. The android apps are converted to images and standardized to maintain uniformity. Processing large malware samples is a challenging task and to address this challenge the android app images are compressed using neural network. The image features are extracted from the dynamic analysis of the dataset. In this paper, we classify and detect variants of malware categories using compressed images and their augmented images using various machine learning models such as k -Nearest Neighbor(kNN), Decision Tree(DT), Support Vector Machine(SVM) and Random Forest(RF). Among these Decision Tree and kNN turns out to perform best, kNN gave an accuracy of $\mathbf{9 8. 8 6 \%}$ over the compressed images and the DT gave an accuracy of $\mathbf{9 9. 5 2 \%}$ over the augmented compressed images. The results show the robustness and adaptability of the DT and kNN to various data transformations making it suitable for classification of malware from android apps.