Malware Classification Method based on Handcrafted Feature Domain

The 21st century has witnessed a major technological boom. Devices once exclusive to large corporations have become available to everyone. This significant development has been accompanied by the use of computers in many aspects of life. Despite the many benefits of these devices, they have led to an alarming spread of malware. There are several methods for detecting and combating such malware, each with its own advantages and disadvantages. In the last decade, software analysis method emerged by converting software binaries into images and then analyzing these images. Similarities in texture patterns were observed across malware families. There are variation in processes of analyzing these images and extracting features from them. Some of them use handcrafted features, while others employ deep learning. Therefore, the main contributions of this paper include presents an in-depth analysis of several works that rely on handcrafted features domain. This analysis covers all aspects of these systems. Then, we propose malware classification method using visualization techniques, where features are extracted based on spatial domain using two types of feature extraction. The proposed method was trained and tested on the Malimg benchmark dataset, achieving a classification accuracy of ${9 9. 1 5 \%}$.