Malicious code classification method based on deep learning

• 1、College of Cyberspace Security, Beijing University of Posts and Telecommunications, Beijing 100876

Abstract：At present, the deep learning network for malicious code classification has a uniform size requirement for malicious code. On the basis of converting malicious code into gray image, the spatial pyramid pooling is used to replace the pooling layer in the traditional network, so that the neural network can accept different sizes of image as input, and combined with the use of depth separable convolution The deep residual module of this paper completes the classification of malicious code image family, and carries out experimental verification on nine different samples of malicious code family. On this basis, the optimal width of malicious code image is discussed. Experimental results show that this method can classify malicious code imagMalicious code classification method based on deep learninges of different sizes uniformly, and has the characteristics of high accuracy and low computation.

Keywords： Malicious code image deep learning classification spatial pyramid pooling