Research On Router Ownership Inference Technology

The router ownership inference technique,which aims to accurately map routers to the Autonomous Systems(AS)they belong to,is an important part of Internet topology research and has important applications in network structure analysis,network boundary identification,network resilience assessment,and inter-domain congestion detection.However,it is difficult to accurately infer the AS to which a router belongs due to the shared address space among neighboring ASes and the existence of third-party addresses.The inferred results of existing methods do not have confidence,while the accuracy still has room for improvement.To this end,this thesis proposes three router ownership inference methods to solve the above problems,focusing on improving the accuracy of router ownership inference,and the main work and innovations are as follows.1.To address the problem that the existing classical router ownership inference method bdrmap IT has insufficient accuracy for routers at the end of probe paths,a router ownership inference method based on the classification of intra-and inter-domain links is proposed.In this method,the distinguished IP address vector distance feature,the AS relationship feature of the IP link and the fan-in and fan-out features are designed to support the discrimination of the IP link type;the probability model with the IP link type as the hidden variable is established for the accurate classification of intra-domain links and inter-domain links;the use of the additional information derived from the link type enriches the basis for router ownership inference,and improves the accuracy of the inference result.Experimental results show that this method can improve the accuracy of router ownership inference.The accuracy rate reaches 96.4% and 94.6%on the two verification sets,respectively,which is 3.2%?11.2% higher than the existing typical methods.2.Aiming at the problem of lack of confidence and insufficient precision in the existing router ownership inference method,a belief propagation-based router ownership inference method is proposed.Based on the target AS router-level topology construction,this method designs four types of co-presence features to support the identification of intra-domain,inter-domain,and extradomain router links for target AS.The probability matrix of co-presence features is constructed by using Markov random field.The belief propagation algorithm calculates the edge probability distribution of router nodes,and obtains the confidence that each router belongs to the target AS.The experimental results show that the method can achieve high-precision inference of router ownership.Its F1 score can reach 91%,which is an increase of 5.6%?8.3% compared with the existing typical methods.The confidence level given in the results is helpful for difference comparison.3.In view of the conflict among the inference results of different methods,a multi-source result fusion method for router ownership inference based on truth discovery is proposed.The method utilizes the idea of truth discovery,takes the results of different methods as input,and performs unsupervised truth discovery through deep neural networks to improve inference accuracy.Experimental results show that the accuracy after fusion of multi-source results is as high as 96.8%,significantly higher than that of existing methods and method using simple voting strategies for fusion.At the end of this thesis,the work of this thesis is summarized,and the future research is prospected.