| 一种基于深度学习的网络流量细粒度分类方法 |
| Deep learning-based method for fine-grained classification of network traffic |
| |
| DOI： |
| 中文关键词: 网络流量分类；深度学习；特征扩展；视频流量；表示学习 |
| 英文关键词:network traffic classification; deep learning; feature augmentation; video traffic; representation learning |
| 基金项目: |
| 作者 | 单位 | | 孔 镇 | 南京邮电大学 通信与信息工程学院，江苏 南京 210003 | | 董育宁 | 南京邮电大学 通信与信息工程学院，江苏 南京 210003 |
|
| 摘要点击次数: 2578 |
| 全文下载次数: 2147 |
| 中文摘要: |
| 随着网络视频服务持续增长，为了更好地区分不同质量的视频服务和管理网络资源，提出了一种基于深度学习的网络流量细粒度分类方法。该方法设计了一种扩展流特征信息的表达方式，将更丰富的特征信息以图片形式表示出来，应用卷积神经网络对图片代表的流数据实现细粒度分类，且省去了特征设计和选择环节。通过对视频数据进行实验，分类准确率达到了93.6%，与文献方法相比，分类性能更好。同时，在其他数据集上的实验结果表明，对网络流量进行粗粒度分类同样能取得很好的分类结果。 |
| 英文摘要: |
| With the continuous growth of network video services, to better distinguish the different qualities of video services and manage network resources, a deep learning-based method for fine-grained network traffic classification is proposed. A feature augmentation is designed by the method in the network traffic for transforming the data information into an image. Using these images representing the corresponding traffic data, a convolution neural network is used for a fine-grained classification (e.g., classifying different quality videos in the same video service), thus omitting the feature design and selection steps. Experimental results made on the collected video traffic dataset show that the accuracy of classification can reach 93.6%, outperforming other existing schemes. Meanwhile, the experiments on other datasets illustrate that the method can achieve good results on coarse-grained classification, such as identifying the type of the traffic. |
| 查看全文 查看/发表评论 下载PDF阅读器 |