基于边权重感知图神经网络的加密流量分类模型OA
EW-GNN:Edge Weight-aware Graph Neural Network for Encrypted Traffic Classification
提出一种面向边权重的图神经网络(edge weight-aware graph neural network,EW-GNN)模型,用于加密流量分类.该模型通过创新的边权重机制,精细化利用图结构信息,有效区分不同边对分类任务的重要性,从而增强对图结构特征的捕捉能力并减少噪声干扰.EW-GNN模型主要由4个核心组件构成:双分支嵌入结构、图神经网络流量表征编码器、交叉门控特征交互机制和端到端分类模块.实验结果表明,EW-GNN模型在ISCX-VPN数据集上的准确率、精确率、召回率和F1分数分别达到94.75％,95.12％,94.83％,94.97％,AUC值为0.954,显著优于其他比较模型.消融实验进一步证实了边权重机制的有效性,启用该机制后模型的各项性能指标均得到超过1.5％的提升.未来工作将集中在拓展模型的应用场景、优化模型结构和训练方法,结合前沿技术提升模型性能,以应对加密流量分类领域的挑战.
This paper proposes an edge weight-aware graph neural network(EW-GNN)model for encrypted traffic classification.By introducing an innovative edge-weighting mechanism,the model effectively leverages graph structural information to distinguish the importance of different edges for classification tasks,thereby enhancing feature extraction capabilities while reducing noise interference.The EW-GNN architecture comprises four core components:a dual-branch embedding structure,a GNN-based traffic representation encoder,a cross-gating feature interaction mechanism,and an end-to-end classification module.Experimental results demonstrate that EW-GNN achieves 94.75％accuracy,95.12％precision,94.83％recall,94.97％F1-score,and 0.954 AUC on the ISCX-VPN dataset,significantly outperforming baseline models.Ablation studies further validate the effectiveness of the edge-weighting mechanism,showing over a 1.5％performance improvement across all metrics when activated.Future work will focus on extending application scenarios,optimizing the model architecture and training strategies,and integrating cutting-edge techniques to address challenges in encrypted traffic classification.
池亚平;白胤廷;杨轩
北京电子科技学院网络空间安全系 北京 100070北京电子科技学院网络空间安全系 北京 100070北京电子科技学院网络空间安全系 北京 100070
信息技术与安全科学
加密流量流量分类边权重图神经网络点互信息阈值过滤机制深度学习
encrypted traffictraffic classificationedge weight graph neural networkpointwise mutual information(PMI)threshold filtering mechanismdeep learning
《信息安全研究》 2026 (6)
533-541,9
评论