Static PE Malware Type Classification Using Machine Learning Techniques
Shao-Huai Zhang, Cheng-Chung Kuo, Chu‐Sing Yang · 2019
In recent years, machine learning techniques have become more and more popular. It is also introduced to the research about malware detection. However, most of research are still focused on binary classification issue, which predicts a file as benign or malicious. Only a small fraction of them work on malware type detection or classification of malware family. This work mainly uses several machine learning models to build static malware type classifiers on PE-format files. A recently released dataset for windows malware detection are used and relabeled into multi-class via VirusTotal, and several efficient and scalable machine learning models are considered. The evaluation results show that our best model, random forest, can achieve high performance with micro avg f1 score 0.96 and macro avg f1 score 0.89, which is better than the model used in referred work.