摘要
为了满足加密移动应用流量识别与管理的需求,针对加密移动应用中使用相似的第三方库而导致同质化流量的问题,提出了一种基于图注意力机制的加密移动应用同质流量识别方法。该方法首先将突发中的多条流相互连接构造流量相关图,在图节点中嵌入数据包特征;其次使用门控循环单元对流量相关图节点内部特征进行学习,使用图注意力神经网络对每个图节点的邻居节点进行上下文流量特征的重要性建模,增强对同质流量中上下文流量的关注与学习,达到识别同质流量的目的。考虑真实场景下存在背景流量与待识别移动应用流量同时存在的情况,在所构建的流量识别模型中加入图节点随机扰动和一致性正则化损失来增强模型的鲁棒性。基于公开数据集,通过与相关研究成果进行实验比较,突出了所提出的方法在加密移动应用同质化流量识别方面的性能优化效果。
For the purpose of traffic identification and management of encrypted mobile applications,a homogeneous traffic identification method for encrypted mobile applications based on graph attention mechanism was proposed to solve the problem of traffic homogeneity caused by the use of similar third-party libraries in encrypted mobile applications.Firstly,the traffic correlation graphs were constructed by multiple streams in traffic bursts,and packet features were embedded in the graph nodes.Then,the gated recurrent units were used to learn internal features of traffic correlation graph nodes.Meanwhile,the graph attention neural network was used to model the importance of the contextual traffic features of neighbor nodes of each graph node,so as to enhance the attention and learning of the contextual traffic in the homogeneous traffic.The proposed method can achieve the purpose of identifying the homogeneous traffic.Considering the coexistence of background traffic and mobile application traffic in the real scenario,the stochastic perturbation of graph nodes and the regularization loss of consistency were introduced to the proposed traffic identification model to enhance the robustness of the model.Experiments were conducted based on the public traffic dataset.The experimental comparison and analysis demonstrat that the proposed method can identify the homogeneous traffic identification of encrypted mobile applications,and provide better performance optimization compared to other approaches.
作者
孙泽沛
王子豪
潘炜
SUN Zepei;WANG Zihao;PAN Wei(School of Computer Science,Northwestern Polytechnical University,Xi’an 710072,China)
出处
《网络空间安全科学学报》
2024年第2期97-106,共10页
Journal of Cybersecurity
基金
国家自然科学基金联合基金项目(U22B2025)。
关键词
加密移动应用
同质流量
流量相关图
图注意力机制
encrypted mobile application
homogeneous traffic
traffic correlation graph
graph attention mechanism
