DeepDroid: Feature Selection approach to detect Android malware using Deep Learning

Abstract

Smartphones are now able to use for various purposes such as online banking, social networking, web browsing, ubiquitous services, MMS, and more daily essential needs through various apps. However, these apps are highly vulnerable to various types of malware attacks attributed to their open nature and high popularity in the market. The fault lies in the underneath permission model of Android apps. These apps need several sensitive permissions during their installation and runtime, which enables possible security breaches by malware. Hence, there is a requirement to develop a malware detection that can provide an effective solution to defense the mobile user from any malicious threat. In this paper, we proposed a framework which works on the principals of feature selection methods and Deep Neural Network (DNN) as a classifier. In this study, we empirically evaluate 1,20,000 Android apps and applied five different feature selection techniques. Out of which by using a set of features formed by Principal component analysis (PCA)can able to detect 94% Android malware from real-world apps.