Malware Detection in Android via Neural Network using Entropy Features
Neeli Khan, Duaa Baig, Muhammad Usman Shahid Khan, Assad Abbas, Mazhar Ali · 2021
Smartphones have become an integral part of everyday life. Android phones dominate the mobile market due to its open architecture. Despite the increased popularity of Android, there has been a significant increase in malicious Android applications in the past few years. Malware APKs can target smartphones through rogue applications to gain control of the devices, for malicious activities. An efficient solution is needed to distinguish malware from benign applications. In this paper, we proposed a detection method based on a deep neural network using entropy features to efficiently detect malware from benign applications. We collected a dataset of balanced 2,000 benign and malicious APKs. We experimented on entropy features for malware detection and on permission features of APKs. Experimental results showed, that our approach achieved 95.56% accuracy in comparison to several machine learning classifiers, such as Random Forest, Naive Bayes etc. Moreover, the proposed methodology performed significantly better on both the permissions and the entropy features.