Research On Network Security Situation Prediction Method Based On Wavelet Neural Network

With internet scale and application scope is expanding,attacks gradually penetrate to network terminal,and attack means have become more sophisticated and covert.Network paralysis,user data leak and other security issues have become increasingly prominent,which brings new challenges to the traditional security defense system with single point and single source.Network security situation awareness is a new technology by scholars at home and abroad in order to solve the problem of single defense.It can realize monitoring and prediction of the entire network security situation,which has important practical significance on adjusting security policies and improving the emergency response capacity.Network security situational awareness includes three key technologies,situation factor acquisition,situation assessment and situation prediction.The former provides data support for the latter.The proposed situation prediction methods are highly dependent on the initial training data,leading to low accuracy of prediction results.Namely,the objectivity of situation assessment and the effectiveness of situation prediction method is the key to influence prediction accuracy.This article launches the research based on the above problems.The specific work is as follows.A situation assessment method based on multi-population genetic algorithm(MPGA)to optimize HMM is proposed.Compared with the traditional static evaluation method,HMM establishes mapping relations between the observation vector and network security situation directly,and it can quantify the risk of network real-time with modeling simple,efficient operation.But the model parameters are the key factors to determine the performance of HMM,this paper uses mixed multi-population genetic algorithm(MPGA)to train the parameters.Through strengthening the information exchange between populations,MPGA can enrich the individual diversity and improve the population average fitness value,so as to improve the algorithm of global search ability and speed up the convergence rate.Finally,the evaluation results are compared with the network attacks,they have good consistence,which indicates that the method is objective and reasonable.A situation prediction method based on Chaotic Particle Swarm Optimization(CPSO)to optimize wavelet neural network(WNN)is proposed.For the situation value has the characteristics of nonlinear time series,the wavelet neural network has a better mapping ability than the traditional neural network to deal with the nonlinear time series data.For the learning algorithm of WNN is sensitive to the initial parameters value,easy to fall into local extremum,this paper uses CPSO that has excellent global and local search features to train the parameters of WNN,to obtain better situation prediction model.Experimental results show that the convergence speed and prediction accuracy of CPSO-WNN are obviously improved.