Intelligent malware classification based on network traffic and data augmentation techniques

Abstract

To prevent detection, attackers frequently design systems to rearrange and rewrite their malware automatically. The majority of machine learning techniques are not sufficiently resistant to such re-orderings because they develop a classifier based on a manually created feature vector. Deep learning techniques like convolutional neural networks (CNN) have lately proven to perform better than more traditional learning algorithms, especially in applications like picture categorization. As a result of this success, CNN network proposed with data augmentation techniques (to enhance the performance) to classify malware samples. We trained a CNN to classify the photos using converted grayscale images from malware files. Our methodology outperforms other methods with an accuracy of 98.80%, according to experimental results.