A Network-wide Traffic Anomaly Detection Method Based on HSMM

Hidden semi-Markov model (HsMM) has been well studied and widely applied to many areas. The advantage of using an HsMM is its efficient forward-backward algorithm for estimating model parameters to best account for an observed sequence. In this paper, we propose an HsMM to model the distribution of network-wide traffic and use an observation window to distinguish DoS flooding attacks mixed within the normal background traffic. Several experiments are conducted to validate our method