Research And Implementation Of BGP Abnormal Event Detection Based On Machine Learning

With the advances in Internet technology,the number of Autonomous System in Border Gateway Protocol(BGP)networks increases rapidly,and the topological structure is becoming complicated.BGP network abnormalities occur frequently.Beacause of the trend of BGP networks,it is more and more important to study the algorithms of abnormal event detection.Among the existing research on BGP abnormal event detection based on machine learning,Most papers study BGP route states in a certain area but few study specific IP prefixes and most study the application of single model algorithm but few study hybrid model.This paper takes IP prefix as the research object and studies the application of hybrid model in the detection of BGP anomalous events and designs two combination model of GBDT combined with LR and GBDT combined with FM for detection.The paper divides the research process into feature data collection,feature processing and abnormal event detection.This paper first calculates the event feature data through the distributed algorithm,removes the unrelated and redundant features through the feature processing,and finally detects the abnormality through the machine learning algorithm.On the basis of the research on the algorithm,the paper design and implement the detection system.The system achieves the detection of abnormal IP prefix and proves that the hybrid model is better than single model in BGP abnormal event detection.