A DNN Approach to Design a Smart Security System for Android Platform

Abstract

Malware is an extensive threat to all computing devices and involves a massive loss for end-users and corporations. This chapter gives a fundamental idea about Android malware types, the intrusion techniques used by malware, the inbuilt security models provided by Android, and the combating techniques used by malware writers to bypass anti-virus detection. Several machine and deep learning approaches have been proposed so far in the Android malware detection field, but most of them have relied on static features due to their lower cost. In this chapter, several experiments are performed using the deep neural network model and the result analysis explains the effectiveness and the limitations of different detection techniques. The experiment with static, dynamic, and hybrid detection techniques achieves an accuracy of 95.40%, 99.66%, and 87.54% respectively. Malware family classification is also conducted using static detection technique and achieved 91.54% accuracy. The content of this chapter provides a methodical way to design an effective detection system.