Research On Network Traffic Anomaly Detection Based On Time-varying Hurst Parameters

With the rapid development of the times,high-tech has gradually integrated into our lives.The Internet has become an indispensable part of our lives.For example,Taobao has replaced traditional shopping methods;Alipay and WeChat have gradually replaced cash.The communication between people has gradually changed from the telephone and text messages of previous years to the current WeChat and Weibo.These technologies are inseparable from the rapid development of the Internet industry.Due to the popularity and popularity of network services,the security of network services has become a major concern of people.The abnormal detection of network traffic will become the key to ensuring the security of network services.Conventional network anomaly detection methods include feature/behavior-based research;statistics-based research;flow-based mining research,etc.Self-similarity analysis of network traffic has become a recent research hot spot.In this study,the time-variant correlation theory is applied to the self-similarity study of network traffic data,and the abnormality detection of network traffic is carried out through modeling and predictive analysis.This thesis first analyzes the model of network traffic and the self-similarity of services,and gives the reasons for the self-similarity of network traffic.The Hurst parameter estimation algorithm is used to describe the self-similarity of network traffic.Secondly,this thesis analyzes the traditional model and points out its shortcomings.It analyzes and studies the model of network traffic,and selects the appropriate network traffic business model for modeling.The FARJ[MA model is simulated.According to the simulation results,the FARIMA model can simulate the network service very well.In order to verify the reliability of the FARIMA model,the ARIMA model and the FARIMA model are used to simulate and predict the randomly generated simulated network traffic data,and the prediction results are compared and analyzed.Finally,the FARIMA model is used to predict the real network data of Bell Labs,and then the abnormal traffic model is predicted.According to the comparison of the prediction results,it is judged whether the network traffic is abnormal,so as to conduct network traffic anomaly detection.Research on network traffic anomaly detection based on Hurst parameter combined with FARIMA model is a new research method of network anomaly,Which is of great significance to the development of network services.