• Study On Optical Communications
  • Vol. 49, Issue 3, 19 (2023)
Qi-feng GUAN*, Su ZHAO, and Xiao-rong ZHU
Author Affiliations
  • Jiangsu Key Laboratory of Wireless Communication, Nanjing University of Posts and Telecommunications, Nanjing 210003, China
  • show less
    DOI: 10.13756/j.gtxyj.2023.03.004 Cite this Article
    Qi-feng GUAN, Su ZHAO, Xiao-rong ZHU. High Precision Traffic Identification Method based on GAN and XGBoost Fusion[J]. Study On Optical Communications, 2023, 49(3): 19 Copy Citation Text show less
    Block diagram of the application identification system
    Fig. 1. Block diagram of the application identification system
    The application identification system
    Fig. 2. The application identification system
    The importance score of features
    Fig. 3. The importance score of features
    The accuracy of the service identification
    Fig. 4. The accuracy of the service identification
    The application identification time
    Fig. 5. The application identification time
    应用类型类型介绍业务类别
    流媒体类视频、直播流量1
    下载类文件类业务的上传与下载2
    即时类微信等社交平台数据交互、网页浏览3
    邮件类型邮箱流量4
    Table 1. The application types
    流量特征字段描述
    dpnum数据流总报文数
    dpl_total数据流报文总长度
    dpnum_s发包率
    dpl_s流发送速率(按字节)
    duration流持续时间
    dpl_max最大报文长度
    dpl_min最小报文长度
    dpl_mean平均报文长度
    dpl_std报文长度标准差
    fpnum上行流报文总数
    fpl_total上行流报文总长度
    fpl_max上行流最大报文长度
    fpl_min上行流最小报文上行长度
    fpl_mean上行流平均报文上行长度
    fpl_std上行流报文长度标准差
    bpnum下行流包总数目
    bpl_total下行流报文总长度
    bpl_max下行流最大报文长度
    bpl_min下行流最小报文长度
    bpl_mean下行流平均报文长度
    bpl_std下行流报文长度标准差
    Table 2. The traffic characteristics
    特征数准确率(%)特征数准确率(%)
    1092.261294.51
    1496.151697.32
    1897.582197.72
    Table 3. The feature number and accuracy of application identification
    业务类型样本数生成样本数总样本数准确率(%)混合样本准确率(%)
    19 9801 00010 98093.7098.85
    26 3242 5328 85690.2097.75
    33 5004 7378 23782.2597.30
    43 1504 8508 00071.5096.98
    Table 4. The number and accuracy of the sample sets
    算法准确率(%)
    朴素贝叶斯44.80
    随机森林80.25
    决策树85.90
    CNN85.38
    循环神经网络82.25
    本文算法97.32
    Table 5. Comparison of accuracy of different algorithm models
    Qi-feng GUAN, Su ZHAO, Xiao-rong ZHU. High Precision Traffic Identification Method based on GAN and XGBoost Fusion[J]. Study On Optical Communications, 2023, 49(3): 19
    Download Citation