Research on network spoofing attack pattern identification based on adversarial machine learning
A network spoofing attack pattern identification method based on adversarial machine learning(AML)is proposed to accurately and automatically identify the behavior patterns of network spoofing attacks and improve network transmission security.In this method,a feature set of network flow tables that can describe the behavior patterns,distribution status and interrelationships of network flow is extracted and input into a generative adversarial network(GAN)for training,so as to construct a network spoofing attack pattern identification model.Under the guidance of the loss function,the generator generates a dataset close to the real sample,and the dataset is then input into a discriminator.The discriminator is designed in a multi-layer structure.The output results of each discriminator are synthesized to obtain the average value,which is taken as the final judgment basis.In combination with the weight matrix,the result is subjected to vote to output the network spoofing attack recognition results.The test results show that the proposed method can extract network flow table features reliably.In addition,its mean absolute error(MAE)percentage for various types of network spoofing attacks is below 0.014 0,with a minimum of only 0.005 8,indicating good identification effect.
国家科技期刊平台
NSTL
万方数据
