基于弹性多尺度熵的网络流量时空特性研究

发布时间：2019-01-05 18:34

【摘要】：随着互联网的不断发展,网络已经成为人们生活中必不可少的部分。如何保证网络安全、稳定、高效的运行成为了当前网络中迫切需要解决的问题。网络流量作为网络中数据流动的载体,对于其时空特性的研究能够帮助了解复杂的网络结构以及网络的动态特性,对于网络的设计、监控等各个方面都有着重要的意义。在理清了国内外多尺度熵理论研究与网络流量特性分析相关工作的基础上,本文基于多尺度熵理论提出了弹性多尺度熵理论,并创新性地将弹性多尺度熵理论用于网络流量时空特性的研究。本文主要研究工作与贡献如下:(1)本文对熵理论以及网络流量特性研究方法进行了系统的阐述,其中熵理论方面对样本熵、多尺度熵、复合多尺度熵理论进行了全面的理论阐述与详细的流程分析。在网络流量特性研究方面对网络流量的获取途径,网络流量的测量,网络流量的特性分析等方面都从理论及应用方面进行了详细的阐述。(2)提出了弹性多尺度熵(Flexible Multiscale Entropy,FMSE)理论,该理论提高了时间序列复杂度量化分析的准确性和稳定性。首先,对现有的样本熵进行了改进,引入了弹性累计的方法。在此基础上,结合当前复合多尺度熵理论的复合计算方法,提出了弹性多尺度熵。最后,基于人工合成的白噪声和1/f噪声序列以及实际振动数据序列对弹性多尺度熵理论进行了验证。实验结果表明弹性多尺度熵比多尺度熵和复合多尺度熵更加准确和稳定,特别是在时间序列较短、尺度因子较大的情况下,弹性多尺度熵在准确性和稳定性上有很大的提升。(3)提出了基于弹性多尺度熵的网络流量时空特性分析方法。首先,对获取的公开网络流量进行多维、分层处理,分别获取不同层次,不同时空维度的网络流量子序列。然后,以本文提出的弹性多尺度熵作为理论工具对网络流量子序列进熵值计算。通过对不同规模的网络进行分层网络流量的复杂性研究,研究发现在整个时间尺度上包序列的复杂度高于流序列的复杂度。字节序列的复杂度整体上高于流序列的复杂度,当网络中有大量小流存在的时候流序列的复杂度会高于字节序列的复杂度。小流对网络层次复杂度有着重要的影响。(4)对本文提出的弹性多尺度熵理论以及基于该理论的网络流量时空特性研究方面的主要工作和贡献进行了总结,并指出弹性多尺度熵以及网络流量研究方面的不足之处,提出了未来的工作方向。
[Abstract]:With the continuous development of the Internet, the network has become an essential part of people's lives. How to ensure network security, stability and efficient operation has become an urgent problem in the current network. Network traffic as the carrier of data flow in the network, the study of its space-time characteristics can help to understand the complex network structure and the dynamic characteristics of the network, for the network design, monitoring and other aspects of important significance. In this paper, the elastic multi-scale entropy theory is proposed based on the theory of multi-scale entropy and the analysis of network traffic characteristics based on the theoretical research of multi-scale entropy at home and abroad. The elastic multi-scale entropy theory is innovatively applied to the study of spatial and temporal characteristics of network traffic. The main works and contributions of this paper are as follows: (1) the entropy theory and the research methods of network traffic characteristics are systematically described in this paper, in which the sample entropy, multi-scale entropy, and multi-scale entropy are discussed in entropy theory. The complex multi-scale entropy theory is described in detail. In the aspect of the research of network traffic characteristics, the way of obtaining network traffic, the measurement of network traffic and the characteristic analysis of network traffic are expounded in detail from the aspects of theory and application. (2) the elastic multi-scale entropy (Flexible Multiscale Entropy, is proposed. FMSE) theory, which improves the accuracy and stability of time series complexity quantization analysis. Firstly, the existing sample entropy is improved and the elastic cumulation method is introduced. On this basis, the elastic multi-scale entropy is proposed by combining the compound calculation method of the current compound multi-scale entropy theory. Finally, the theory of elastic multi-scale entropy is verified based on synthetic white noise, 1 / f noise sequence and actual vibration data series. The experimental results show that elastic multi-scale entropy is more accurate and stable than multi-scale entropy and composite multi-scale entropy, especially in the case of shorter time series and larger scale factor. Elastic multi-scale entropy greatly improves its accuracy and stability. (3) an analysis method of time-space characteristics of network traffic based on elastic multi-scale entropy is proposed. Firstly, multi-dimensional and hierarchical processing of the obtained open network traffic is carried out, and the sub-sequences of network traffic of different levels and different space-time dimensions are obtained respectively. Then, the elastic multi-scale entropy proposed in this paper is used as a theoretical tool to calculate the progressive entropy of network traffic subseries. It is found that the complexity of packet sequence is higher than that of flow sequence in the whole time scale by studying the complexity of hierarchical network traffic in different size networks. The complexity of byte sequence is higher than that of stream sequence on the whole, and the complexity of stream sequence is higher than that of byte sequence when there are a large number of streams in the network. Small-stream has an important influence on the hierarchical complexity of the network. (4) the main work and contribution of the elastic multi-scale entropy theory and the research of space-time characteristics of network traffic based on the theory are summarized. The inadequacies of elastic multi-scale entropy and network traffic research are pointed out, and the future work direction is proposed.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP393.06

【参考文献】