Malware Traffic Classification Based on Multidimensional Features Learning

In the domain of network security, techniques for detecting and classifying malware based on network traffic play a pivotal role. These techniques can effectively predict and identify potential malicious activities through precise analysis of network behavior data, thereby providing essential protection for information systems. Noting the limitations of existing malware traffic classification methods in extracting sequence features and the insufficiencies of statistical features in capturing the full spectrum of traffic behaviors, we introduces an innovative approach. Our study maps network communication traffic onto a graph-based structure, leveraging the properties of graphs to uncover behavioral differences among various types of malware. By conducting a detailed analysis of the spatial attributes of nodes and edges within the graph, combined with sequence information, we propose a classification method for malware traffic based on multidimensional features. Furthermore, we have compiled multiple datasets from diverse environments and performed a series of experiments. The results indicate that the proposed method enhances both the accuracy and efficiency of malware classification.