Network behaviour anomaly detection using Holt-Winters algorithm

Because security attacks in IP networks are becoming ever more difficult to detect, we must develop better detection systems to protect network users. One possible method which could be used is forecasting network behaviour. In this paper, we study the Holt-Winters forecasting algorithm in aberrant behaviour detection. We have built a test platform which collects real IP network traffic and calculates forecasts for collected network traffic. Open source software NfSen is used to produce graphs about collected and forecasted network traffic to the end user.