Research On Network Security Situation Awareness Technology Based On Neural Network

In recent years,with the continuous development of information technology,the scale of network has been gradually expanded,and network application has penetrated into all aspects of society.However,at the same time,the network threats are constantly increasing,and the incidents of network attacks also occur from time to time,and traditional border-based security protection measures cannot well grasp the trend of network security.In view of this problem,this paper combines the application advantages of neural network in network security,and conducts research from both situation assessment and situation prediction to improve the ability of network security situation perception and prediction.Firstly,in view of the problems of low detection accuracy and high false positive rate in DNN attack detection,this paper proposes an attack detection model which combines batch normalization and deep neural networks.The model adds a batch normalization layer to the hidden layer of the deep neural network,optimizes the output of the hidden layer,and uses Adam's adaptive gradient descent optimization algorithm to automatically optimize the parameters of BNDNN to improve the model's anomaly detection capability.By analyzing the impact of different network attacks on the security situation,a situation assessment index based on attack factors is constructed.This method uses the situation of being attacked to reflect the current situation of the host,and then comprehensively calculates the situation value of the entire network based on the situation information of all hosts in the network And quantify the situation level,and realize the perception of network security situation.Then,by analyzing the timing characteristics of network security situation,combining the security situation values obtained above,the Long Short-Term Memory(LSTM)is used to solve the situation prediction problem;at the same time,the excellent learning ability and global search ability of genetic algorithm(GA)is used to optimize the network structure parameters of LSTM,and establishes a situation prediction model based on GA-LSTM to realize the prediction of network security situation.Finally,in order to verify the effectiveness of the method,this paper uses NSL-KDD data set to conduct a situation assessment experiment and compares it with SNN,KNN,DNN and other models to prove the feasibility and effectiveness of the situation assessment model.At the same time,this paper also collected the real-time network data,using the real network data to verify the situation assessment model validity again;The situation prediction was performed with GA-LSTM through the situation data after the situation assessment,and compared with the standardized LSTM network and time series analysis method.The results prove the effectiveness of the prediction model in this paper.