The Improved Model for Anomaly Detection Based on Clustering and Dividing of Flow
Ao Liu, Bin Sun · 2019
To avoid the serious phenomena of the computing congestion and the invalid sampling redundancy, an improved anomaly detection model of data link layer based on analyzing for the characteristics of network flow is proposed, which is focus on the computing optimization and the sampling reduction in each period by an improved feedback mechanism. Based on the approximate optimal number of clustering calculated, the time points of feedback can be allocated dynamically instead of a unified one. And the quantization of flow characteristics is also obtained by analyzing its period division of time coordinate. The processes of improved mechanism and algorithm are designed, and the simulation experiments verify that the model proposed can improving the efficiency of anomaly detection to a large degree, without reducing its accuracy.