Malware classification using gray-scale images and ensemble learning

• Published in: 2016 3rd International Conference on Systems and Informatics (ICSAI) pp. 1018 - 1022
• Main Authors: Liu Liu, Baosheng Wang
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.11.2016
• Subjects: Bagging; Deep learning; Feature extraction; Gray-scale; Image coding; machine learning; Malware; malware classification; Malware detection; n-gram; Training
• DOI: 10.1109/ICSAI.2016.7811100

Malware authors can easily generate obfuscated and metamorphic malware to evade the detection of anti-virus software using automated toolkits. To identify these variations, we design an efficient system to detect and classify the malicious software. We construct a controlled disassembly files from malware. And we convert disassembly files into gray-scale images. In order to improve the efficiency, we use the local mean method to compress gray-scale images, which are mapped into feature vectors. To classify malware, we propose a novel ensemble learning which is based on K-means and the diversity selection. Finally, our experiments show that our method is able to effectively classify the malware.