基于TVCSE拟合的网络入侵预测方法

Prediction Method of Network Intrusion Based on Tvcse Fitting

针对传统算法网络入侵预测中,网络检测数据相关性不强,大规模网络入侵导致预测准确性不高的问题。提出了一种基于时序变化率曲面极值(Time-series variation curved surface extremum:TVC-SE)拟合的网络入侵预测算法。对网络流量信息数据进行实时获取,以时序为时间度量构建变化率曲面模型;分别对时序变化率曲面模型的局部极值进行迭代计算,将共有的局部极值作为最终的预测极值进行存储,同时以该时刻的协同局部极值为参考,提高了网络入侵预测的准确性。仿真实验表明,该测试方法能够达到较高的测量精度,虚警率比传统算法平均降低了12.4%,预测时间减少在2.5 s左右,在不增加成本,符合网络程序计算复杂度的情况下,满足了网络入侵预测的要求。

In view of the traditional algorithm in network intrusion prediction, the network test data correlation is not strong, large-scale network invasion in prediction accuracy is not high. Is put forward based on the temporal variation curved extre?mum (Time-series variation curved surface extremum:TVCSE) fitting network intrusion prediction algorithm. Access to the network detection data in real time, based on sequential time metrics build rate surface model;Of temporal variation surface model of local minima for iterative calculation, the total of the local extremum of as the final prediction of extreme value for storage, at the same time to the moment of coordinated local extremum for reference, improved the accuracy of net?work intrusion prediction. Simulation experiments show that the test method can achieve higher measuring precision, virtu?al measurement rate than the traditional algorithm was reduced by 12.4%on average, forecast time reduced at around 2.5 s, without any increase in cost, network program under the condition of computational complexity, meet the requirements of the network intrusion prediction.