Network Traffic Anomaly Detection Method Based On Features Of Catastrophe Series Theory

To detect network anomaly can discover many existing malfunctions and performance problems, which becomes an efficient way to tackle network problems. To enhance the capability to detect network malfunctions as well as performance problem is very meaningful to improve the network availability as well as reliability, and guarantee the quality of service. The network traffic anomaly is referred as the current traffic is much different from the normal one, then decreasing the network performance. The traffics, causing network anomaly to happen, includes the useless traffic induced by network attacks, users'unsuitable use towards network resources and network congestion caused by malicious attack from clients. Besides, malfunctions of network instruments and links can cause network anomalies too. The network anomalies can seriously affect network performance, cause network congestion and even network interruption.Currently, at home and abroad, the anomaly detection methods seen network traffic as a time sequence, generally can not describe the anomaly flow of network with the structure of the dynamics of non-linear, non-smooth, complexity and the Catastrophe characteristic. View of the current network traffic anomaly detection by the existing problems, according to network traffic is often driven by multiple factors of control, their behavior will show non-linear, non-smooth and complexity, and the nature of its internal evolution equation of power will show the characteristics of mutation. With self-similar way, space reconstruction methods and statistical physics methods, calculated the macro features of network traffic volume, which can choose to reflect significant changes in network traffic unusual features as control variables, proposes a network traffic anomaly detection method based on the catastrophe series theory model.The experimental results show that proposed network traffic anomaly detection method has low false alarm ratio very high accuracy. This paper explores the research on network security mechanisms and is provided with important studying and practice significance.