Unveiling encrypted traffic types through hierarchical network characteristics
Yufan Chen,Jiahai Yang,Susu Cui,Cong Dong,Bo Jiang,Yuling Liu,Zhigang Lu
DOI: https://doi.org/10.1016/j.cose.2023.103645
IF: 5.105
2023-12-22
Computers & Security
Abstract:The wide adoption of encrypted traffic brings challenges to network management. Previous studies propose different approaches to tackle this problem. However, most of them still struggle to strike a balance in three dimensions, interpretability, robustness and efficiency. As countermeasures, we propose using a hierarchical feature approach to enhance the comprehensiveness behavior description of network behavior to achieve the balance of the three points. Instead of aggregating under-layer features, we have employed the collector mechanism to directly generate the hierarchical features. This approach can overcome the issue of inaccurate feature generation caused by multiple concurrent session flows. To measure the generalization ability persuasively, six datasets are reorganized to simulate the scenario with unknown applications. Besides, different statistical models are applied with state-of-the-art features to conduct experiments. Experiment results show that the proposed method can achieve the F1 score above 98% on the reorganized sets, which demonstrates the effectiveness of the proposed hierarchical approach.
computer science, information systems