Visual classification method for malicious codes

The invention relates to the technical field of malicious code classification, in particular to a visual classification method for malicious codes. According to the technical scheme, the method comprises the following steps of S1, data preparation: acquiring a malicious code data set and converting a binary file into a grey-scale map, S2, feature extraction: extracting image features by using a backbone network based on Resnet, and S3, setting an MA module: adopting the MA module of multi-scale convolution operation to perform dimension reduction on a feature map, reduce the calculation amount and extract local features, so as to obtain the malicious code data set. S4, a CA module: carrying out the weight distribution of a feature map channel through employing an attention channel module; and S5, carrying out the classifier training and prediction. According to the method, a new malicious code visualization data set is constructed on the basis of a BODMAS data set, a good data basis is provided for subsequent