Malicious Code Family Classification Method Based on Spatial Pyramid Pooling and Deep Residual Network
Jianyi Liu, Yansheng Qu, Jiaqi Li, Yunxiao Wang, Jing Zhang, Hongshan Yin · 2021
Malicious code and its derivative code have become a major threat to network security. At present, some methods transform malicious code into images and use deep learning to classify families. However, these family classification methods based on deep learning has a problem that the malicious code images need to be uniformly scaled before model training, which may result in the loss of potential malicious code image features. This paper proposes a malicious code classification network based on spatial pyramid pooling and deep residual network. The network can accept malicious code images of any size as input, and solve the problem that neural network input requires uniform image size. The experimental results show that the classification accuracy of this paper is 99.09%, and the recall is 96.69%, which is 2% higher than other methods on the same dataset.