Research On Network Security Situation Element Acquisition And Prediction Technique

With the development of networks,the network architecture becomes more and more complex.The overall security requirements for the network are becoming more stringent.How to ensure the security of the network is an urgent problem to be solved.Network security situational awareness technology can dynamically understand the security of large-scale network.It established a complete and unified security situation system from large-scale network.Extracting the situation elements analyze the security performance of the whole network.It can provide security strategy to predict the dynamic trend of security risk for the whole network.As a result,this technology can be better adapted to the security needs of networks.The thesis mainly aims at the network security situation element acquisition and the situation forecast two aspects to carry on the research,the concrete work is as follows:Firstly,in order to speed up the convergence speed of the deep belief network and improve the acquisition accuracy of the situation elements under the condition of small sample.A deep belief network security situation element acquisition mechanism based on batch normalization is proposed.On the one hand,the integration of batch normalization in the deep belief network to solve the gradient disappearance problem stable network training;On the other hand,in the deep belief network output layer,an improved active learning algorithm is proposed to reverse fine-tune the deep belief network.It actively chooses training samples in each iteration to balance sample types.Theoretical analysis and experimental data simulation results show that this mechanism can solve the problem that the convergence speed of deep neural network is too slow.The gradient disappears and the classification of small class samples is accurate.In the acquisition accuracy,convergence speed and algorithm complexity is superior to the not improved depth of belief network situation element acquisition mechanism.Secondly,in order to improve the accuracy of network security situation prediction,the generative adversarial network is used to simulate the development process of the situation.The situation prediction is realized from the time dimension.A network security situation prediction mechanism based on differential Wasserstein generative adversarial network is proposed.In the process of situation prediction based on generative adversarial network,in order to solve the problem of network difficulty training and gradient instability in generative adversarial network,a loss function using Wasserstein distance as generative adversarial network is proposed.A difference item is added to the loss function to improve the classification accuracy of situation value.Experimental results and theoretical analysis show that this mechanism has some advantages in convergence,prediction accuracy and complexity compared with other mechanisms.