Texture-Based Malware Family Classification
Nitish Kumar, Toshanlal Meenpal · 2019
Malware is one of the major threats on internet whose count is increasing rapidly every year in millions. Most of the time similar malware files are modified for creation of new variants and most of the existing technique are obfuscated. So, malware visualization using image helps to overcome this problem. Motivated by the visual similarities of malware, this paper presents the malware family classification using some texture-based features. HOG (Histogram of Oriented Gradient), LBP (Local Binary Pattern) and GIST features are used as feature descriptor. We have performed our experiment on 6370 samples of 15 different families. We have observed the highest accuracy of 98.73% using HOG descriptors by SVM classifier. This method shows better results than existing malware classification approaches.