摘要
针对现有匿名网络流量识别模型准确率低的问题,提出了一种基于一维残差卷积神经网络的Tor匿名网络流量识别模型。该模型根据网络流量各特征之间相互独立,无内在关联的特性,采用一维卷积进行特征提取,并采用最大池化筛选、保留关键特征,通过引入跳跃连接解决深层网络存在的退化问题,降低训练时梯度消失的风险,使得模型可进一步加深,提高识别准确率。实验结果表明,该模型优于常用的SVM、KNN、ResNet等对比模型,将Tor匿名网络流量识别准确率提高至98.87%,具体匿名应用类型识别准确率提高至96.14%。
Aiming at the problem of low accuracy of existing anonymous network traffic identification models,this paper proposed a tor anonymous network traffic identification model based on one-dimensional residual convolutional neural network.According to the characteristics of network traffic are independent of each other,no intrinsic correlation,the model uses one dimensional convolution for feature extraction and maximum pooling to filter and retain key features.At the same time,by introducing skip connections to solve the degradation problem of deep networks and reduce the risk of gradient disappearance during training,the model can be further deepened to improve the recognition accuracy.Experimental results show that the model is better than commonly used comparative models such as SVM,KNN,ResNet,etc.,improves tor anonymous network traffic identification accuracy to 98.87%and specific anonymous application type identification accuracy to 96.14%.
作者
冯松松
王斌君
Feng Songsong;Wang Binjun(School of Information Network Security,People's Public Security University of China,Beijing 100038)
出处
《现代计算机》
2022年第17期1-8,41,共9页
Modern Computer
基金
国家社会科学基金(20AZD114)。
关键词
TOR
匿名网络
暗网
流量识别
卷积神经网络
Tor
anonymous network
dark web
traffic identification
convolutional neural network
