Research On Frequent Episodes Mining Of Time Series For Network Security Situation Prediction

Network security is more and more serious and has been a very serious problem now. Faced with the large-scale network security time-series data, to solve the situation prediction of network security efficiently and accurately has very important research significance, and will help network administrators to make decision analysis and prevention measures.Network security situation prediction will implement the trend prediction of the future network security situation, based on accurate network security situation awareness and sufficient network security situation understanding. Large-scale network security situation prediction mainly depends on the effective mining of massive security series data. The existing network security situation prediction method can predict the future situation value of unit time, and then according to the trend of the continuous value of unit time to assess network security situation. And situation values of unit time are often influenced by disturbance, and difficult to predict. This paper attempts to directly predict a future period of time trend change of network situation, according to historical time-series data. And not have to carry on situation analysis on the value of the moment. Namely the paper is based on trend analysis to research network security situation prediction of time-series data. In order to solve this problem, based on time-series data to predict the trend of network security, the main researches of this paper are as follow:(1)This paper presents a neural network method based on time-series data, to predict network security situation trend change in a future period of time. According to the structure characteristics of the time-series data and the theory of prediction method based on the trend, the time-series data of network security are segmented with each segment representing a trend, and converted to sample set suitable for network situation prediction, using sub-curve fitting. The sample set is used to implement network security situation trend prediction based on BP neural network. Using the Honeynet data, the paper completes the simulation experiment, to verify the effectiveness and applicability of network security situation prediction method based on the trend.(2)This paper Mainly studies the mining method for predict of frequent episodes in events sequences, to implement network security time-series trend prediction. Firstly, according to the statistical regularity of network security events, network security time-series data are preprocessed to implement the segmentation, discrete events in sub-segments, and are transformed into events sequences. Then introducing the related concepts of frequent episodes in events sequences implement the mining for predict of frequent episodes in time-series data. Considering network security situation prediction would be sensitive to the relative occurrence time of events of frequent episodes, the paper introduces a limit to the length of event occurrence to achieve the improvement of existing mining algorithm of frequent episodes, to extract the needed frequent episodes for prediction, which would be used to predict network security situation trend change in a future period of time.