Loading
Preliminary Exam Talks
Alidade: A Big-Data Approach to Geolocation
|
Abstract
Geolocation, the process of locating an IP address on the globe, has remained a key research challenge for a long time. With the unprecedented growth in use of online services, viz., Facebook, Foursquare, Twitter, that implicitly or explicitly require the mapping between IP address and location, improving the accuracy of geolocation systems is gaining a lot of traction. Although there has been extensive research in this domain, the focus has been predominantly on using active probes from a number of vantage points, chosen strategically, to improve the accuracy of estimates. We present Alidade, a system that attempts to geolocate the entire routable IP address space, while absolutely refraining from employing active probes. Alidade’s hypothesis is that with a huge corpus of data, it is feasible to design a geolocation system that generates good estimates, without any active probing. To this end, the system is designed to assimilate a wide variety of datasets, both measurements-based and heuristics-based, none of which have been collected specifically for Alidade. To the best of our knowledge, Alidade represents the first academic work on a passive geolocation system that estimates en masse, the locations of all IP addresses in its input datasets. At the heart of Alidade, is an iterative constraint satisfaction algorithm that refines the location estimates of addresses, each time it runs; our design also borrows few of the techniques introduced by Octant.
Advisor(s): Bruce Maggs
Jeffrey Chase, Landon Cox, Walter Willinger