Deep Learning Methods Based Malware Classification

Abstract

In recent years, malicious software has become a major threat and one of the serious security problems affecting not only computer users but also all technological processes and devices that operate in all areas of our lives. In addition, malware that is increasing and getting stronger with each passing day becomes difficult to detect automatically with existing methods and causes the continuous development of new security methods. Detection and classification methods and many tools using these methods have been developed to protect computer systems and deal with malware. In this article, deep learning-based methods created with iterative neural networks, gated iterative units and attention mechanism to automatically classify these malware using API call sequences of 8 different malware types are proposed and compared. The methods developed in this study were tested on the dataset created by collecting the datasets used in the literature and preprocessing them to be used in the methods. The results obtained showed that recurrent neural networks gave good results in malware detection, and the results are given in the article. © 2022 Elsevier B.V., All rights reserved.