On the Evaluation of AS Relationship Inferences

The commercial contractual relationship among Autonomous Systems (ASs) is important for understanding the reachability and traffic flow in the Internet. Several heuristic algorithms have been proposed for inferring AS relationships by extracting information from Border Gateway Protocol (BGP) routing tables. However, there is no evaluation of the AS relationships inferred from existing algorithms. In this paper, we present the methodology for obtaining partial AS relationships, and use them to evaluate these algorithms. Our result shows that although existing algorithms achieve good overall accuracy, the accuracy on the peer-to-peer relationships is quite poor. Furthermore, we propose an algorithm for inferring AS relationships by taking advantage of this partial information. We show that our proposed algorithm outperforms existing algorithms for inferring peer-to-peer relationships by 1.8 3.8 times in accuracy. In addition, the overall accuracy of our algorithm is higher than that of existing algorithms