Network Security Situation Prediction Based On Multiple Model Weights Extraction And Fusion

In order to improve the traditional security methods can not be timely to deal with the current situation of network threats. In this article, the situational awareness technology is integrated into the network security, predicting the security situation of the network environment in the future b y using the situation prediction technology in situation awareness, it is advantageous for managers to prepare against threats in advance, and in time to reduce the losses caused by threats. At present, the current research on situational awareness is not mature especiall y the situation prediction is in the initial stage. Although some situation prediction methods have been proposed, the uncertaint y of the network security event limits the prediction accurac y of the traditional forecasting methods.Therefore, this article mainly does several aspects work as follow.First of all, in view of the old data used in the research on network securit y situation prediction, to find a new network data source.By normalizing the data published on the CNCERT website, we get the value of the network securit y situation in the mainland of China in the 52 week of 2015, and using this set of network security situation value to carr y out the research work in this paper. Secondly, in view of the problem of poor real-time performance and low accurac y of existing situation prediction methods, a trend prediction method based on multiple model weights extraction and fusion is proposed, which can realize the combination forecasting of multiple models. Due to different t ypes of situation prediction methods have different advantages and disadvantages, we comprehensively consider the range of the application of the each model, prediction accurac y and other characteristics, the selection of the gra y prediction model, BP neural network prediction model, Elman neural network prediction model and RBF neural network prediction model for fusion. The establishment of four kinds of forecasting methods focuses on the network security situation prediction by the single forecasting model and the final combination forecasting model. Finally, we analyze the error through the simulation experiment, comparing the prediction outcoms of each individual model and the combined model, the contrastive anal ysis results show that the combination forecasting method is more accurate than the single forecasting method.