Master's Defense

IP Geolocation Aggregation

Speaker:Xiao Xiao
xx at cs.duke.edu
Date: Monday, July 2, 2012
Time: 4:00pm - 6:00pm
Location: D344 LSRC, Duke
Romit Roy Choudhury, Xiaowei Yang

Abstract

IP geolocation queries, which ask for returning the physical location of a query IP, are widely studied because of a wide range of applications such as online advertising, location-based access restrictions, context-aware security and cloud computing. In the current geolocation methods, some IP addresses have no direct measurement data and no geographic hints in the reverse DNS name. We therefore study aggregation, a method to make reasonable predictions for those IP addresses. Aggregation combines predictions for other IP addresses in the longest prefix match that contains at least one other address for which we have a prediction already. We look into several aggregation approaches to determine which gives us the best results. An evaluation of aggregation using host parser shows that the average error size can be as small as 16 kilometers.
Advisor(s): Bruce Maggs