基于神经网络的大数据云平台DDoS攻击检测方法研究  

Research on neural network based DDoS attack detection method for big data cloud platform

作　　者：葛菁[1] 赵巍[1] 徐亦丹[1] GE Jing;ZHAO Wei;XU Yidan(Institute of Technology,East China Jiaotong University,Nanchang 330100,China)

机构地区：[1]华东交通大学理工学院,江西南昌330100

出　　处：《现代电子技术》2020年第24期102-104,共3页Modern Electronics Technique

基　　金：教育部产学合作协同育人项目(201901149002);江西省教育厅科学技术研究项目(GJJ171486);南昌市移动通信重点实验室(2018⁃NCZDSY⁃008)。

摘　　要：针对基于OpenFlow协议的云平台的安全性问题,文中对分布式拒绝服务攻击(DDoS)检测方法进行研究。通过引入自组织映射(SOM)神经网络,利用网络流量的包数、速率、生存周期等特征建立网络的输入特征向量对SOM中的输入层、竞争层进行合理的优化,借助Stacheldraht工具生成网络的训练和测试数据。在实验时,文中基于不同的流量数据集训练得到3个不同的SOM网络。测试结果表明,所提方法对于恶意流量的识别准确率可达98%以上,误判率可降低至0.5%以下,证明了神经网络在DDoS攻击检测中的可用性。In allusion to the security problem of cloud platform based on OpenFlow protocol,the detection method of distributed denial of service(DDoS)attack is researched.With the introduction of the self⁃organizing mapping(SOM)neural network,the input eigenvector of the network is established by means of the characteristics of the number of packet,rate and life cycle of network traffic to reasonably optimize the input layer and competition layer in the SOM neural network.The training and testing data of the network are generated with Stacheldraht.In the experiment,three different SOM networks were obtained on the basis of training of different traffic datasets.The testing results show that the recognition accuracy of this method for malicious traffic can reach more than 98%,and its misjudgment rate can reduce less than 0.5%,which proves the availability of neural network in DDoS attack detection.

关 键 词：云平台 DDoS攻击监测 自组织映射 神经网络 特征优化 流量检测 实验测试 

分 类 号：TN911.7-34[电子电信—通信与信息系统] TP311[电子电信—信息与通信工程]