Malware and Benign Detection Using Convolutional Neural Network

Abstract

Malware is a virus program written to enter a system and damage or alter files and data. A computer virus is more vulnerable as it makes changes or deletes files and can enter into a system as an attachment of images or audio and video files. It also infects the system through the downloads on the Internet. This paper proposes an intelligent learning model to recognize malware using convolutional neural network approach. Here, malware detection identification models through binary classifier, and two approaches of deep neural network classification and convolutional neural network are used. A database consisting of 600 exe files of which 300 malware files and 300 benign files have been considered here for building the malware and benign detection model. All the exe files are converted into images files. The contemporary CNN-based binary classification which takes the grayscale images as input. The convolutional neural networks-based classification model proves accuracy of 93% in discriminate from malware and benign files. The convolutional neural network-based malware detection model has higher performance when compared with deep neural network classification model trained with GIST features of image.