Research On Network Security Situational Awareness Based On Machine Learning

With the acceleration of networking and digitalization,the means of cyber attacks are becoming more and more diverse,and the issue of cyberspace security has attracted more and more attention.The current passive defense strategies based on firewalls and IDS can no longer meet the needs of the current complex network environment.Therefore,research on network security situational awareness,which is a representative of active defense technologies,is of great research and exploration value.To this end,this paper analyzes and summarizes the existing network security situational awareness models,and conducts in-depth research on them based on network security situational awareness and machine learning related technologies.The main work and innovations of this paper are summarized as follows:(1)Propose a network security situation assessment method based on PRFREFCV algorithm for feature selection,and optimize Light GBM parameters using GA.In this paper,after using the PRF algorithm to calculate the importance of each feature,the recursive elimination algorithm with cross-validation is used to select the optimal feature subset,avoiding the influence of redundant features on downstream algorithms,and improving the evaluation efficiency and accuracy of the model.At the same time,the GA algorithm is used to globally optimize the important hyperparameters of Light GBM,avoiding the influence of subjective factors,and improving the accuracy and generalization ability of the model.Experimental results prove that the accuracy of the situation assessment method reaches more than 99%,which is higher and more efficient than the traditional network security situation assessment algorithm.(2)Propose a network security situation prediction method based on KPCA feature reduction and PSO-RBF neural network.In this paper,we use KPCA algorithm to calculate the cumulative contribution rate and eliminate redundant features to achieve the goal of feature reduction,reducing the computational burden of the base model.Furthermore,we use PSO algorithm to optimize the parameters of RBFNN,avoiding the issue of falling into local optimal solutions and improving prediction accuracy.In addition,we introduce the concept of inertia weight into random weight to further improve the global optimization ability of the algorithm.Through experimental comparative analysis,the proposed situation prediction model shows good prediction accuracy.Through simulation and comparison experiments,the situation assessment and prediction model proposed in this paper can effectively reflect the current and the near future security status of the network,and have positive significance for the development of network security situation awareness in the future.