A Research of Approximate Entropy’s Clustering Analysis in the Detection of Abnormal Flow

How to detect mixed or unknown network attacks is difficult in anomaly detection .Network traffic has an important feature of nonlinear dynamics, The paper proposes a new method to detect abnormal traffic by approximate entropy, one of the important dynamics parameters, the relevant parameters are tested and compared in experiments. Finally,sequence of approximate entropy generated are processed by cluster analysis to improve accuracy, the results show that the method could identify traffic with mixed or unknown attacks well. Furtherlly improvements of the method are made a discussion.