Malware Detection Using Artificial Neural Networks

Abstract

This paper deals with improvement of malware protection efficiency. The analysis of applied scientific researches devoted to creation of malware protection systems suggests that the improvement of mathematical tools using modern neural network models based on deep neural networks is a promising trend in the development of malware detection systems. Also, the results of analysis have determined the need to create a development method for the deep neural network architecture suitable for use within the modern malware detection means. As part of the study, a method for developing a deep neural network architecture designed to detect malicious software has been suggested. In contrast to the existing methods, it helps avoid long-term numerical experiments to determine the expediency of application of the neural network model and optimize its structural parameters during the development. At the same time, multiple experiments conducted using Microsoft BIG-2015 malware database have shown that the method constructs a neural network model that provides a detection error commensurate with the error of modern malware detection systems. Prospective research is related to the adaptation of the suggested method for the application of deep neural networks in behaviour analysers.