Android Malware Detection using Convolutional Deep Neural Networks

Abstract

Deep learning in general and convolutional architectures, in particular, have pushed the limits of the current state of the art in the field of computer vision and the processing of natural languages and speech. Recently, these techniques have been applied to detect mobile malware and have once again shown their ability to remedy this type of problem. However, the most suitable deep network architecture for malware detection remains an open issue. In this paper, we investigate the possibilities of convolutional neural networks for efficient detection of mobile malware. Specifically, we address the impact of using inception based and multichannel architectures on network performance. We achieve an accuracy of 92% using a multichannel model on a set of 50000 malware and 50000 benign applications.