Applying RNN and J48 Deep Learning in Android Cyber Security Space for Threat Analysis

Abstract

Recurrent Neural Networks (RNN) are a special class of deep learning algorithms using neurons or nodes, and have received much attention in the subject of data science in the recent years. In RNN, the input nodes take into consideration not only the current inputs, but the previously perceived outputs as well – hence the term recursive. In today’s context, smartphones are very much a part of almost every individual’s daily lives. The demand, development and usage of Android devices is massive. As Android devices dominate the current market share, the question of security naturally arises in our complex world. Consequently, the amount of malware data available for research is voluminous as well. This publication demonstrates the power and efficiency of RNN applied onto Android malware data. We study a procured dataset, with over 4000 entries labeled as malicious or benign. From our experiment and data analytics, we present a prediction accuracy of 0.964 using RNN.