A HIERARCHICAL CLASSIFICATION METHOD FOR TYPICAL ANONYMOUS NETWORK TRAFFIC

To address the challenge posed by anonymous networks in concealing user identities for network security management, this paper proposes a hierarchical classification method for typical anonymous network traffic. Statistical features of data flows were extracted and combined with machine learning for coarse classification. Time-related features and raw packet byte features from coarsely classified traffic were then fused for traffic reconstruction, followed by fine-grained classification using deep learning. Experiments on four typical anonymous network applications demonstrate accuracies of 99.70%、98.47%、and 96.05% in identifying anonymous traffic, types, and user behaviors, respectively. The proposed method exhibits enhanced flexibility and superior classification performance compared with existing approaches.