Research of Anomaly Detection Based on Time Series

With the continuous deterioration of the network environment, a variety of viruses, Trojans continue to affect the security of the network. Through the network traffic anomaly detection and analysis can efficiently find problems existing in the network. This paper discusses the network traffic flow data predict and network anomaly detection, network traffic prediction using ARMA model, network anomaly detection using the exponential smoothing model. ARMA model supplies the expectation value to abnormal detection, at the same time exponential smoothing model can restoration historical flow data, making the following traffic forecast more accurate. A network traffic predict and network anomaly detection system has been developed, with which can find network anomaly and send alarms, thus improve network stability and robustness.