...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Journal of Real-Time Image Processing >Deep learning-based real-time VPN encrypted traffic identification methods
【24h】

Deep learning-based real-time VPN encrypted traffic identification methods

机译：基于深度学习的实时VPN加密流量识别方法

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

With the widespread application of virtual private network (VPN) technology, real-time VPN traffic identification has become an increasingly important task in network management and security maintenance. Since traditional encrypted traffic identification technology is not effective in feature extraction and selection, this paper proposes two deep learning-based models to classify the traffic into VPN and non-VPN traffic, identify VPN traffic generated by six different applications much further. Our models utilize convolutional auto-encoding (CAE) and convolutional neural network (CNN), respectively, preprocessing the traffic samples into session pictures, to accomplish the experiment objectives. The CAE-based method, utilizing the unsupervised nature of CAE to extract the hidden layer features, can automatically learn the nonlinear relationship between original input and expected output. The CNN-based method performs well in extracting two-dimensional local features of images. Experimental results show that our models perform better than traditional identification methods. In the two-category identification, the best result comes from the CAE-based model; the overall identification accuracy rate is 98.77%. Among the six-category identification, the best result comes from CNN-based model; the overall identification accuracy rate is 92.92%.
机译：随着虚拟专用网（VPN）技术的广泛应用，实时VPN流量识别已成为网络管理和安全维护中越来越重要的任务。由于传统的加密流量识别技术无法有效地进行特征提取和选择，因此本文提出了两种基于深度学习的模型，将流量分为VPN和非VPN流量，进一步识别了六个不同应用生成的VPN流量。我们的模型分别利用卷积自动编码（CAE）和卷积神经网络（CNN），将流量样本预处理为会话图片，以达到实验目的。基于CAE的方法利用CAE的无监督性质提取隐藏层特征，可以自动了解原始输入与预期输出之间的非线性关系。基于CNN的方法在提取图像的二维局部特征方面表现良好。实验结果表明，我们的模型比传统的识别方法性能更好。在两类识别中，最好的结果来自基于CAE的模型。总体识别准确率为98.77％。在六类识别中，最好的结果来自基于CNN的模型。总体识别准确率为92.92％。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号