基于Copula理论的网络资源风险预测研究

发布时间：2018-06-28 16:09

本文选题：Copula + 网络性能指标　；参考：《湖北工业大学》2017年硕士论文

【摘要】：随着云计算和海量存储以及新的网络技术的急速发展,网络的安全与维护面临着许多的问题。首先,网络规模日趋庞大,网络带宽增长迅速,现有的网络监控设施缺乏足够的监测手段和扩展性来应对这种变化。其次,网络应用越来越丰富,以及各类潜在于流量中的网络威胁,使得网络资源变得匮乏,导致网络性能不断下降乃至崩溃,我们必须尽早预测出风险并展开适当的风险评估,进而保护网络资源。因此,实时有效地对网络资源风险进行预测及评估具有非常重要的研究意义。本课题就网络资源风险预测及评估展开了研究。和以往的网络资源风险预测方法不同,以往通常是针对于某一个指标进行单独的预测及评估,而本文选取了多个网络资源指标,提出了一种基于藤Copula-GARCH模型的预测方法,并采用贝叶斯网络对网络风险展开了适当的评估。论文首先研究了基于Copula理论的网络资源风险预测的方法。根据多个网络资源指标,分别进行相关数据收集,经归一化处理后结合GARCH模型求出各自的边缘分布函数以及边缘密度函数,进行了相关的统计分析。接下来,结合Copula理论和藤结构,通过秩相关系数来确定藤的逻辑结构,运用pair-Copula分解模型进行了建模,构建出多元变量联合分布函数的计算模型,并采用蒙特卡洛仿真对网络资源风险VaR进行了求解。此外,论文在网络资源风险预测的基础上探讨了运用贝叶斯网络进行风险评估的方法。该方法根据网络拓扑结构和已收集的网络资源数据的统计分析,计算出各指标的先验概率,并通过BNT工具箱求出对应的后验概率,最后得到各网络资源指标对网络风险的影响度排序以及风险等级,通过影响度排序分析出最有可能引起网络发生风险的网络资源指标。实验结果表明,基于Copula理论的网络资源风险预测的方法,根据Copula将变量相依结构与边缘分布分开描述的优点,结合GARCH模型和藤结构将多元联合分布分解成二元copula函数对,简化了计算,同时根据预测值与给定阈值的比较结果,运用贝叶斯网络进行网络资源风险评估,具有良好的评估效果,达到了预期的目的。
[Abstract]:With the rapid development of cloud computing, mass storage and new network technology, network security and maintenance are facing many problems. First of all, the scale of the network is becoming larger and larger, and the network bandwidth is growing rapidly. The existing network monitoring facilities lack sufficient monitoring means and scalability to cope with this change. Secondly, more and more network applications, and various potential network threats in traffic, make network resources become scarce, leading to network performance decline and even collapse, we must predict the risk as soon as possible and carry out appropriate risk assessment. Then protect the network resources. Therefore, it is very important to predict and evaluate the risk of network resources in real time. This topic has carried on the research to the network resources risk forecast and the appraisal. Different from the previous methods of network resource risk prediction, it is usually used to predict and evaluate a certain index separately. In this paper, a prediction method based on Fujiao Copula-GARCH model is proposed by selecting several network resource indicators. Bayesian network is used to evaluate the network risk. Firstly, the paper studies the method of network resource risk prediction based on Copula theory. According to several network resource indexes, the related data were collected, and the edge distribution function and the edge density function were calculated by normalized processing and GARCH model, and the related statistical analysis was carried out. Then, combining Copula theory and rattan structure, the logical structure of rattan is determined by rank correlation coefficient, and the pair-Copula decomposition model is used to model the calculation model of joint distribution function of multivariate variables. Monte Carlo simulation is used to solve the network resource risk VaR. In addition, based on network resource risk prediction, the paper discusses the method of risk assessment using Bayesian network. Based on the statistical analysis of network topology and collected network resource data, the priori probability of each index is calculated, and the corresponding posteriori probability is obtained by BNT toolbox. Finally, the influence degree of each network resource index on the network risk and the risk grade are obtained. Through the influence degree ranking, the network resource index which is most likely to cause the network risk is analyzed. The experimental results show that the method of network resource risk prediction based on Copula theory, according to the advantages of Copula to describe the dependent structure of variables and the edge distribution separately, combined with GARCH model and rattan structure, the multivariate joint distribution is decomposed into binary copula function pairs. The calculation is simplified and the Bayesian network is used to evaluate the network resource risk according to the comparison between the predicted value and the given threshold value. It has a good evaluation effect and achieves the expected purpose.
【学位授予单位】：湖北工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP393.08

【相似文献】