Image-based malware classification using convolutional neural network

Abstract

In this paper, a malware analysis method that analyzes images learned by artificial intelligence deep learning to enable protection of big data by quickly detecting malware, including ransomware, is proposed. First, more than 2,400 datasets frequently used by malware are analyzed to learn and image data with a convolutional neural network. Data are then converted into an abstract image graph and parts of the graph extracted to find the group where malware exist. Through comparative analysis between the extracted subsets, the degree of similarity between these malware is analyzed experimentally. Fast extraction is achieved by using deep learning. Experimental results obtained indicate that use of artificial intelligence deep learning can enable fast and accurate malware detection by classifying malware through imaging.