Hybrid multi-scale modeling of network traffic

This paper presents a hybrid multi-scale modeling of network traffic based on wavelet transform, Poisson process and fractal geometry. It demonstrates that network traffic has statistical properties and self-similarity at different scales. Both sharp time behavior and long-range dependence of network traffic can be well simulated by our model. A classic algorithm is improved to calculate the Hurst parameters and a new algorithm is also presented to calculate the Holder index based on the space characterization properties of wavelet basis. Comparing to former works our model provides several more parameters to characterize network traffic. The efficiency of this model is verified by using DARPA Intrusion Detection Evaluation dataset. The result shows that our model is effective to simulate network traffic.