| As the rapid development and application of wireless network, the scale of wireless network is becoming larger and larger, the application is becoming more and more complicated, the task of management is becoming more and more heavily. Wireless network have the characteristics of wireless transmission medium, dynamic topology, lack of supervision and so on, So that it is vulnerable to attack and results in abnormal network events. In order to realize reliable data transfer and reasonable internet resource distribution, it is important to comprehend the control mechanism and complicated behavioral character of wireless network. Traffic prediction and anomaly detection has significant meanings for management, layout and design of large scale wireless network.. In this paper, the research is focus on improving wireless network traffic prediction accuracy through the establishment of prediction model, and proposing a wireless network security event detection mechanism.Firstly, this paper narrates the character and architecture of wireless network comprehensively ,describing the current research at home and abroad, it offers the foundation for the following researches.Secondly, this paper analyzes several main characters about wireless network traffic in actual wireless network environment and introduces several typical wireless network traffic prediction model.Thirdly, an ARMA prediction model is established. At first, this model pre-smooth wireless network traffic, and then determine the order of model by using AIC criterion. Estimated parameters are resolved in a dynamic real-time way. The result shows that ARMA model has higher prediction accuracy than AR model and MA model. Compared with the multi-step prediction, the result of ARMA model in the case of one-step can reflect the trend of real wireless network traffic better.Fourthly, wireless network traffic prediction techniques apply to network security incidents. First getting prediction result by using network traffic prediction model and comparing the difference of actual and prediction result, so that a wireless network security incidents detection mechanism is established. Network simulation using DDoS attack as a example show that it achieves a higher detection rate and low false positive rate. |
