A novel method of network traffic anomaly detection

The purpose of this paper is to propose a new algorithm which is based on combining the linear model and the method about smooth exponential. It is used to simulate the AR model of the Single sliding window sequence of observations and gets the average value of the square of slide window observed values´ residual noise and achieve the previous statistics on the latter statistic forecast with exponential smoothing method so as to decide whether the network traffic is normal or not. This algorithm is more efficiently comparing with GLR method and more reliable comparing with smooth exponential method, and has been proved is effectively in the detection for network traffic anomalies.