BiTCN malware classification method based on multi-feature fusion
Bona Xuan, Jin Li, Yafei Song · 2022 International Conference on Image Processing, Computer Vision and Machine Learning (ICICML) · 2022
At present, there are a large number of variants of malware and are constantly updated. The generalization of detection algorithms based on traditional machine learning methods is attenuated and the accuracy rate is constantly declining. This paper proposes a multi-feature fusion bi-directional temporal convolutional networks (BiTCN) malware classification algorithm. According to the n-gram sequence features of malware opcode operand and application programming interface (API), the algorithm introduces BiTCN to mine bidirectional time series features, enhances the algorithm’s time series feature extraction ability, and improves the classification accuracy of the algorithm. The experimental results show that BiTCN can effectively avoid sequence features easily interfered by obfuscation technology, and can effectively detect malware variants.