Adaptive Method For Anomaly Detection Based On Kalman Filter

Nowadays, with the rapid development of computer network technology, the node number, scale and topology complexity of Internet have been increasing exponentially. However, at the same time, network attacks, P2P flood and computer virus are coming along the way, which brings about anomaly network flow resulting in network congestion, denial of service and even network paralysis.In this paper, I make a deep analysis on the variance behavior of a network device caused by anomaly and thus improve the original Kalman Filter by exponentially smoothing its noise vector in a adaptive way.In addition, I put forward an adaptive kind of algorithm and model for anomaly detection which is real-time, expansible and predictable. By implementing this low-cost and real-time algorithm, we can detect anomaly on single-node or multi-node in a network of large scale.Finally, I validate the algorithm and model in the real network by using SNMP and also make certain analysis on the data I get.