Research On The Detection Model Of Anomaly Network Traffic

Network traffic burst anomaly refers to the significant abnormal changes suddenly appeared in the network traffic, So timely detection about traffic anomalies in network has great importance for rapid positioning abnormal and adopting follow-up corresponding measures. However, Traffic anomaly detection algorithm is required to analyze and deal with massive real-time network business data accurately along with the scale and speed increase continuously in networks.Aim at questions about large storage space requirements and quite complex calculation are exist in calculating the wavelet coefficients with standard discrete wavelet transform, wavelet detection technology which has been existed is improved in this thesis according to the rapid upgrading of wavelet theory to calculate the coefficient. The standard method relies on convolution of the original signal with FIR filter structures. Fast lifting wavelet transform adopts CDF 9/7 filter, Java has been chosen for its platform independent character.which proposal is mainly aim to break the original filter form and formate a series of scale structures. Achieve its computional complexity is only half of the original convolution filter methods without increase in the extra storage. In thesis, potential function was quoted to achieve flow despcrition in abnormal changes in the networks, and studied catastrophe characteristics in network traffic with catastrophe theory, here also proposed detection model to describe anomalies in traffic. Combinate network characteristics and based on the existing and common network traffic anomaly detection model and method, this paper proposed potential function which is used in network traffic flow anomalies. Together with the existing static and dynamic detection algorithm, this function can get network traffic anomaly detection model.This thesis designed network anomaly detection model and verified accuracy and real-time of model by comparing with other detection model in experiments. Detailed design of the model and test flow and test case were given in thesis.