Network Traffic Detection Based On Fuzzy-wavelet Analysis

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, researchers have proposed many network detection methods based on network traffic. Such methods include threshold detection, exponential smoothing detection and generalized likelihood detection methods. However, those methods are endowed with such deficiencies: a large number of history data are required; detection probability is low; and generalized likelihood detection method requires much computing complexity. Aiming at these problems, this paper puts forward a network security system based on network traffic. Based on the randomicity the uncertainty, we mainly adopt the fuzzy lookup traffic prediction method according to the fuzzy membership functions from history data; on this basis, the network anomaly detection algorithm is put forward on the basis of prediction errors, which is aiming at detecting network anomaly; at last, according to the self-similarity and fuzzy decision making, a method of categorizing network attacks is put forward, which is to determine the type of network attacks.The experimental results show that proposed network traffic prediction method has very high accuracy; and the anomaly detection method based on prediction error is fast in detection speed, high in detection probability, and requiring very small number of data. This paper explores the research on network security mechanisms and is provided with important studying and practice significance.