Implementation Of Clustering Analysis Based BGP Routing Abnormal Detection System

Autonomous systems'(ASes)instability can affect the accessibility and reliability of network services.Abnormal monitoring and detection to the autonomous systems is helpful to optimize the network quality of service,and locate network service failure.Because of the complexity of the global autonomous systems' routing topology,routing data quantity is extremely large.To detect stable abnormal accurately and fast,specific algorithm and system architecture should be designed and implemented.Previous studies have failed to solve the accurate abnormal detection of autonomous systems and the real-time detection of global autonomous systems at the same time.By using the similarity between autonomous systems dynamic analyzing clustering characteristics is helpful to further mining the correlations between each node in interdomain routing system.This study adopted the wavelet method to transform time series of AS updates,and proposed a new metric to measure the similarity between ASes,and specific clustering algorithm.Further,based on the proposed clustering algorithm and the routing updates data,carried on the cluster analysis to autonomous domain.And applied clustering-based AS abnormal prediction model(CAAP)on this dataset.Then,found out the relationships between ASes by clustering analysis and predict AS' abnormal dynamic by watching on a part of ASes rather than the whole system.Finally,designed and implemented a CAAP-based real-time detection system(CRDS),which is consists of data collector,CAAP abnormal analysis,abnormal prediction,visualization and configure module.The test results show that the CAAP-based abnormal detection scheme can accurately predict AS anomalies.Using CRDS to monitor the global abnormal events of ASes is meaningful to save computational and storage overhead.