Inferring geography from BGP raw data

Research to date has analyzed the Internet AS-level topology at a worldwide level of detail. Every inference found for an AS is extrapolated from the global set of AS paths gathered from public monitors, independently of the geographic location of the ASes. This approach is useful when the Internet is analyzed at a very coarse level. However, it may be misleading if the analysis is more focused on a specific geographical region. The risk is that the particular characteristics that the Internet has in that region may be lost. An AS connection that has been identified in a global analysis may hide multiple connections located in different geographic regions, each with its own characteristics. Moreover, a couple of ASes may establish different economic relationships in each geographic region where they are connected. In this paper we propose an innovative technique to geolocate the AS connections retrieved from BGP raw data, in order to highlight the Internet regional characteristics. The analyses that we performed revealed some regional characteristics, in terms of graph properties and inter-AS economic relationships, that should be taken into account in a future analysis of the Internet.