Visual malicious code classification based on target detection algorithm

With the rapid development of computer technology, the number of malware gradually increases. To address the problem of malware family classification, the paper combines visualization techniques with target detection algorithms to construct a malware classification model. The malware is visualized as grayscale images, and the similar texture features of the grayscale images are used to represent the similar code structure of the same family of malware. The Yolov4 network is constructed and data enhancement techniques are used to train the grayscale image data. The experimental results show that the model has an average accuracy of 91.5% for malicious code classification, which is higher than traditional machine learning KNN and random forest.