MCSMGS: Malware Classification Model Based on Deep Learning

Abstract

As a major threat to cyber security, malware has been increasingly damaging national security. This paper proposes a malware classification model, i.e. MCSMGS model (Malware Classification Based on Static Malware Gene Sequences), that combines the static malware genes with deep learning methods. The model extracts the malware gene sequences that have both material attribute and informational attribute. Then it makes distributed representation for each malware gene to represent the intrinsic correlation and similarity. Finally, the SMGS_CNN (Static Malware Gene Sequences – Convolution Neural Network) module is used to construct the neural network to analyze the malware gene sequences and realize malware classification. The experimental results show that the classification accuracy is greatly improved and up to 98% with the MCSMGS model. CNN model is more effective than the traditional SVM model.