Abstract: The studies on internet measurement have facilitated the development of router-level topology discovery， while subnets in the network layer provide a more detailed intermediate complementary view. In order to deal with the low accuracy caused by insufficient subnet boundary conditions and completeness， a multi-characteristic subnet discovery algorithm was proposed. The characteristics of the traceroute path of IP in the same subnet were studied， and were then combined to generate more precise subnet boundary determination conditions. By filtering the completeness of a subnet， the search space of the candidate subnet was narrowed， and the problem of subnet discovery was solved iteratively. The experimental results show that the proposed algorithm can discover subnets more accurately than other existing algorithms， reduce false positive rate， and improve efficiency. Finally， subnets were inferred on six geographically disperse ISPs， and the common subnet characteristics appearing in these ISPs are analyzed.

Key words: topology discovery, traceroute, subnets discovery, multiple characteristics, topology analysis