-
摘要:
针对机载网络高度动态、高度不稳定造成流量监测设备难以在有限的监测周期内完成完整数据流负载特征的提取，限制了基于深度学习的流量分类方法的应用问题，提出了一种鲁棒性增强的机载网络流量分类方法。通过数据预处理及缺失样本处理方法将数据流映射为灰度矢量集合，基于完整的数据流训练数据集实现鲁棒性增强的长时递归卷积神经网络（RE-LRCN）分类模型的训练，在线上分类阶段，通过分类模型实现样本缺失数据流负载空间特征及数据流时序特征的提取，并进行数据流分类。通过在数据包缺失的流量测试数据集上的实验结果表明，所提方法可以有效抑制数据包缺失对分类准确性能的恶化。
Abstract:The highly dynamic and highly unstable characteristics of the airborne network make it difficult for traffic monitoring equipment to extract the complete data flow load characteristics within a limited monitoring period, thus limiting the application of the deep learning based traffic classification method. Aimed at this problem, a robustness-enhanced airborne network traffic classification method is proposed. First, data stream samples are mapped to gray vector sets by data preprocessing and missing sample processing methods. Then, the Robustness-Enhanced Long-term Recursive Convolutional neural Network (RE-LRCN) classification model is trained based on the complete traffic training set. Finally, in the online classification stage, the loading space features of packets-sample deficient data flows and timing features of data flows are extracted and the traffic is classified with the RE-LRCN model. The experiment results on the packets-sample deficient test set show that the proposed method can effectively suppress the deterioration of the accuracy of classification due to the missing of packet samples.
-
Key words:
- airborne network /
- traffic classification /
- deep learning /
- feature extraction /
- robustness
-
图 3 低样本缺失敏感的LSTM网络结构
Figure 3. Structure of LSTM network with low sample loss sensitivity
图 5 不同分类模型下测试数据集的分类准确性能对比
Figure 5. Comparison of classification veracity performance under different classification models in Te_Set
图 6 样本缺失条件下分类模型训练及测试过程
Figure 6. Training and testing process of classification model with loss of samples
图 7 样本缺失条件下不同分类模型整体分类准确率对比
Figure 7. Comparison of overall classification precision among different classification models with loss of samples
图 8 样本缺失条件下不同分类模型类可信度与类覆盖度对比
Figure 8. Comparison of precision and recall among different classification models with loss of samples
表 1 AN_Set中数据流样本分布
Table 1. Distribution of data flow samples in AN_Set
类别 数量 比例/% A 3 113 25.53 B 3 671 30.10 C 2 265 18.57 D 3 146 25.80 合计 12 195 100 
下载: 导出CSV
表 2 数据流负载空间特征提取网络的结构
Table 2. Structure of data flow loading spatial feature extraction network
层数 类型 输入 窗口大小 窗口数量 步长 边缘填充 输出 1 卷积层+激活函数 (28×1)×1 3×1×1 3 1 边缘一致 (28×1)×3 2 最大池化层 (28×1)×3 3×1×3 1 3 边缘一致 (10×1)×3 3 卷积层+激活函数 (10×1)×3 3×1×3 6 1 边缘一致 (10×1)×6 4 最大池化层 (10×1)×6 3×1×6 1 3 边缘一致 (4×1)×6 5 全连接层 (4×1)×6 10 10×1
下载: 导出CSV
表 3 实验相关参数设置
Table 3. Experimental parameter setting
参数 数值 输出灰度矢量尺寸 28×1 前部子流截取窗口大小 28 负载空间特征提取网络训练轮次e1 50 负载空间特征提取网络学习率η1 0.5 时序特征提取网络训练轮次e2 50 时序特征提取网络学习率η2 0.3 惩罚因子衰减参数γ 0.5 惩罚因子上调参数ϕ 0.2
下载: 导出CSV
-
[1] 霍大军.网络化集群作战研究[M].北京:国防大学出版社, 2013:66-68.HUO D J.Operation of network swarm[M].Beijing:National Defense University Press, 2013:66-68(in Chinese). [2] 梁晓龙, 何吕龙, 张佳强, 等.航空集群构型控制及其演化方法[J].中国科学:技术科学, 2019, 49(3):277-287.LIANG X L, HE L L, ZHANG J Q, et al.Configuration control and evolutionary mechanism of aircraft swarm[M].Scientia Sinica(Technologica), 2019, 49(3):277-287(in Chinese). [3] CAO X B, YANG P, MOHAMED A, et al.Airborne communication networks:A survey[J].IEEE Journal on Selected Areas in Communications, 2018, 36(9):1907-1926. doi: 10.1109/JSAC.2018.2864423 [4] 吕娜, 杜思深, 张岳彤, 等.数据链理论与系统[M].北京:电子工业出版社, 2018:7-9.LV N, DU S S, ZHANG Y T, et al.Theory and system of data link[M].Beijing:Publishing House of Electronics Industry, 2018:7-9(in Chinese). [5] 梁一鑫, 程光, 郭晓军, 等.机载网络体系结构及其协议栈研究进展[J].软件学报, 2016, 27(1):96-111. http://d.old.wanfangdata.com.cn/Periodical/rjxb201601005LIANG Y X, CHENG G, GUO X J, et al.Research progress on architecture and protocol stack of the airborne network[J].Journal of Software, 2016, 27(1):96-111(in Chinese). http://d.old.wanfangdata.com.cn/Periodical/rjxb201601005 [6] NGUYEN T T, ARMITAGE G.A survey of techniques for internet traffic classification using machine learning[J].IEEE Communications Surveys & Tutorials, 2009, 10(4):56-76. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=e768100b27445c060ed19cc662b2e2ee [7] SHAFIQ M, YU X, LAGHARI A A, et al.Network traffic classification techniques and comparative analysis using machine learning algorithms[C]//20162nd IEEE International Conference on Computer and Communications (ICCC).Piscataway: IEEE Press, 2016: 2451-2455. [8] 李勤, 师维, 孙界平, 等.基于卷积神经网络的网络流量识别技术研究[J].四川大学学报(自然科学版), 2017, 54(5):959-964. doi: 10.3969/j.issn.0490-6756.2017.05.011LI Q, SHI W, SUN J P, et al.The research of network traffic identification based on convolutional neural network[J].Journal of Sichuan University (Natural Science Edition), 2017, 54(5):959-964(in Chinese). doi: 10.3969/j.issn.0490-6756.2017.05.011 [9] 王勇, 周慧怡, 俸皓, 等.基于深度卷积神经网络的网络流量分类方法[J].通信学报, 2018, 39(1):14-23. http://d.old.wanfangdata.com.cn/Periodical/txxb201801002WANG Y, ZHOU H Y, FENG H, et al.Network traffic classification method basing on CNN[J].Journal on Communications, 2018, 39(1):14-23(in Chinese). http://d.old.wanfangdata.com.cn/Periodical/txxb201801002 [10] SHI H, LI H, DAN Z, et al.An efficient feature generation approach based on deep learning and feature selection techniques for traffic classification[J].Computer Networks, 2018, 132:81-98. doi: 10.1016/j.comnet.2018.01.007 [11] WEI W, SHENG Y, WANG J, et al.HAST-IDS:Learning hierarchical spatial-temporal features using deep neural networks to improve intrusion detection[J].IEEE Access, 2018, 6(99):1792-1806. [12] ZOU Z, GE J, ZHENG H, et al.Encrypted traffic classification with a convolutional Long short-term memory neural network[C]//2018 IEEE 20th International Conference on High Performance Computing and Communications.Piscataway: IEEE Press, 2018: 329-334. [13] TONG W, ZHEN H, WEI W, et al.Early-stage internet traffic identification based on packet payload size[J].Journal of Southeast University(English Edition), 2014, 30(3):289-295. http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=dndxxb-e201403006 [14] 范竣翔, 李琦, 朱亚杰, 等.基于RNN的空气污染时空预报模型研究[J].测绘科学, 2017, 42(7):76-83. http://d.old.wanfangdata.com.cn/Periodical/chkx201707013FAN J X, LI Q, ZHU Y J, et al.A spatio-temporal prediction framework for air pollution based on deep RNN[J].Science of Surveying and Mapping, 2017, 42(7):76-83(in Chinese). http://d.old.wanfangdata.com.cn/Periodical/chkx201707013 [15] GREFF K, SRIVASTAVA R K, KOUTNÍK J, et al.LSTM:A search space odyssey[J].IEEE Transactions on Neural Networks & Learning Systems, 2016, 28(10):2222-2232. http://d.old.wanfangdata.com.cn/Periodical/jdq202011006 [16] DONAHUE J, HENDRICKS L A, ROHRBACH M, et al.Long-term recurrent convolutional networks for visual recognition and description[J].IEEE Transactions on Pattern Analysis & Machine Intelligence, 2014, 39(4):677-691. https://arxiv.org/abs/1411.4389 [17] LIU J, SHAHROUDY A, XU D, et al.Spatio-temporal LSTM with trust gates for 3D human action recognition[C]//European Conference on Computer Visio.Berlin: Springer, 2016: 816-833. -
下载:
点击查看大图
计量
- 文章访问数: 540
- HTML全文浏览量: 70
- PDF下载量: 288
- 被引次数: 0
