Malicious code detection method based on image segmentation and deep residual network RESNET

Many existing malicious code detection methods based on deep learning basically have high accuracy, but when detecting malicious code families with high similarity, due to the lack of obvious training features, the detection accuracy is seriously reduced. To solve this problem, this paper proposes a malicious code detection method based on image segmentation and deep residual network. Firstly, the original gray image is transformed into more distinctive sample data by image segmentation technology, which makes the data set increase the distance between classes and reduce the distance within classes, and then the feature extraction and training are carried out through the deep residual network. In the paper, Malimg data set is used to test. Compared with the sample data set without image segmentation technology, the detection accuracy is improved from 95.86% to 98.94%, and the detection accuracy of similar malicious code family is increased from 51.85% to 81.48%