Research On Network Security Situation Assessment And Prediction Method Based On Neural Network

In the current complex network environment,malicious attacks emerge in endlessly,and network security issues become increasingly prominent.At this time,network security situation assessment and prediction provide comprehensive support for the overall network security concept,timely monitoring and dynamic tracking of network security status.To this end,this dissertation conducts an in-depth study of the network security situation assessment and prediction methods.The main work and innovations are as follows:Construct a multilayer and dimensionality network situation assessment index system.In this dissertation,the network security situation is divided into threat sub-situation,fragile subsituation and basic operational sub-situation for assessment,and the bottom-level situation index quantification formula is proposed,which effectively solves the problem of single dimension of situation factors.In order to visually describe the hidden information among the situation elements in the index system,this dissertation proposes a network security situation assessment method based on Inception-CNN,and constructs a situation assessment model integrating situation factor extraction and situation analysis.In this dissertation,Convolutional Neural Network(CNN)is combined with the improved Inception module to take into account the global and detailed features of the situation.In addition,high-boost filtering is introduced to increase the model's sensitivity to different sub-situations,and multiple situation estimators are constructed to obtain the correlation between the overall situation and sub-situations,so as to realize the quantitative analysis of the network security situation.Aiming at the time series characteristics of the situation prediction data,a network security situation prediction method based on PSO-MCELman is proposed,realizing the construction of dynamic prediction model with a cycling internal and external feedback.By adding a memory layer and a concentration layer to the traditional Elman neural network structure to generate a multi-step delay operator,an improved MCElman(Memory Concentration Elman)neural network is formed to achieve long-term memory of external information.And the cumulative distribution function is used to assign additional time weights to the input data of the concentration layer.At the same time,the initial weights and self-feedback gain factors of MCElman neural network are optimized by using the Particle Swarm Optimization(PSO)algorithm to solve the problem that the neural network is easy to fall into local optimum and to achieve accurate prediction of network security situation.The CIC-IDS 2017 data set was used to validate the above two methods.The experimental results show that the situation assessment model based on Inception-CNN has superiority in accuracy,and can obtain the sub-situation types that have the greatest impact on the situation,which provides a new idea for the in-depth analysis of situation assessment.At the same time,through multiple sets of comparative experiments,it shows that the proposed prediction model is more efficient and accurate for network situation prediction.