City-level Geolocation Based on Routing Feature

For the problem that traditional approaches for IP geolocation based on delay measurement are difficult to apply to network with weak connectivity such as China's Internet, in this paper, we utilize its hierarchical topology and proposed an approach of City-level geolocation based on routing feature. Taking IPs with known geographical location as reference nodes, this approach extracts identifying IPs of candidate regions or cities based on decision tree learning algorithm. We match the path of the target with IPs above called identifying features, and then select the region or city whose identifying feature is contained on the target's path as geolocation result. This approach improves the average accuracy for the Internet with weak connectivity hierarchical topology to 93% vs. 73% for the previous learning-based geolocation approach.