LandmarkMiner: Street-level Network Landmarks Mining Method for IP Geolocation

Ruixiang Li, Rui Xu, Yuanyuan Ma, Xiangyang Luo

Abstract: High-confidence network landmarks are the basis of IP geolocation. However, existing landmarks acquisition methods had weakness such as high time cost and insufficient landmarks number. For this, LandmarkMiner, a street-level network landmarks mining method, is proposed based on service identification and domain name association. First, LandmarkMiner trains classifiers using the scanning results of IPs with known hosting service type, identifies the hosting service type of target IPs using the trained classifiers, and obtains the classified IPs’ domain names using DNS. Then, according to institutional names, a database associating institutional name with possible domain names is built by statistical relationship, which is obtained between the known institutional names and their domain names. Finally, geographical location of IP's domain name after classification is matched in the database and online maps, thereby obtaining landmarks and evaluating reliability of them. LandmarkMiner has mined 9,423 reliable street-level landmarks from 304M IPs in 18 cities. Comparing with existing methods, LandmarkMiner increases the number of reliable street-level landmarks significantly and can be applied in different network connectivity conditions.