Real-time detection method for network traffic anomalies

The traditional network management tools usually generate alarms according to the preset threshold. This method is straightforward, but it has poor adaptability. Therefore, the technique to detect network anomaly appeared, sometimes it can find even predict faults of the network. A novel method to detect network anomaly is introduced. According to the method, the observation series of network traffic is transformed and assumed to be stationary, an AR model is built to fit it, and a statistic is defined for network anomaly detection. The result shows that the proposed method has more advantages compared with GLR method.