Evaluation Research Of Network Security Situation Awareness Based On PSO And Neural Network

In recent years,the Internet has developed rapidly and our life has changed dramatically.The increase of the scale of the network and the environment of the network has become more and more complex,at the same time,the technology of network attack technology is constantly updated.The new type of attack tools is emerging,such as DDOS,Ransomware,Parton trojans invasion,etc.The traditional network security technology is often helpless for these attacks;therefore,the network security problem appears more and more serious.The concepts of network security situational awareness have emerged in recent years,the main idea is to extra,understand and evaluate the intrusion information,and then predict the future network security behavior,it is significant to study the situation awareness technology.This paper expounds the background and research significance of the network security situation awareness and introduces the traditional technology.The present technology is introduced at home and abroad to analyze the problems and challenges.The basic concept of network security situation awareness is introduced.The main purpose of multi-source data fusion is to fuse multi-source heterogeneous security data which is detected by multi-sensor detection and used for situation assessment.Situation assessment is the core of situational awareness and a qualitative and quantitative description of network security situation.This paper presents cross-layer particle swarm optimization with adaptive mutation to improve the traditional D-S evidence and evaluate the network security situation.Experimental results show that this method can evaluate the network security situation better.According to the security data of the situation assessment,the network forecast adopts the corresponding prediction technology to predict the development trend of the security situation.An improved RBF neural network is proposed to predict the network security situation.RBF neural network in the practical training process easily appear basis function selection and large amount of data problems,therefore,this paper introduces a method combining Fuzzy C-means and Hierarchy Genetic Algorithm to improve the learning process of traditional RBF neural network.This method mainly determines the number of neurons through the algorithm of Fuzzy C-means,and then use the hierarchical genetic algorithm to gain the neural network center and width,finally,use the least square method to calculate the weight of the hidden layer to the output layer and obtain accurate prediction value.