Anjum believes the TULIP Geolocation application can be improved significantly. At least there are few ideas that we can try. For this, either a group of undergraduate students or an active masters student is required. The resultant work can easily be the thesis of masters level.
- See http://www.slac.stanford.edu/comp/net/tulip/. Basically TULIP uses pings to a target from landmarks at known locations and converts the minimum RTTs to estimate the distances. Then uses the distances with mulitlateration to estimate the location of the target
- To improve TULIP one needs the right selection of landmarks, i.e. good (working landmarks) at the right locations (not too far from the target), straddling the target, and with a a reasonable estimate of the indirectness (directivity or alpha) of the path from the landmark to the target (so we can reasonably accurately estimate the distance). One also needs a reasonable density of landmarks (e.g. number of targets/100,000sq km)
- The landmarks come from PingER and perfSONAR sites. We have a reasonable density in the US, Pakistan and Europe. Currently Anjum is getting better than 20km accuracy for Pakistani targets
- As the number of landmarks goes up so does the accuracy, but so does the time to make the measurements (pings).
- One needs to find the optimal density
- Anjum proposes to speed up the measurements using a cluster for parallelization and also proposes to improve the adaptation of alpha based region. He regards the adaptive geolocation and parallelization as MS projects.
Skills: Perl/Python, statistical analysis, RESTful interfaces
Mentor: Anjum Naveed UM Kuala Lumpur, Les Cottrell SLAC
References: Adaptive Geolocation of Internet Hosts, to be published, Abstract:
- IP based geolocation is a widely used geolocation technique because of its ability to geolocate the hosts where GPS or other techniquws become ineffective or unavailable. Measurement-based geolocation techniques utilize landmarks to make end-to-end delay measurements and compute the host location based on delay to distance mappings. Fewer landmarks and/or inaccurate delay to distance mapping leads to large error margins. In this research an Adaptive-IP-Geolocation (AIG) technique is proposed. AIG is based on the detailed analysis of a large amount of fine grained delay measurement data collected using PingER and PerfSonar worldwide deployments. Based on the analysis of this data, AIG divides the world into regions. A two tier approach is used where tier one landmarks identify the region of a target host. This is followed by geolocation of the target host using regional landmarks only. A variable alpha is introduced for delay to distance conversion. The relationship between alpha and the RTT values is empirically computed for each region using ground truth from real delay measurements. Results show that AIG outperforms previous techniques with the error margin reduced to 25 km or less for the majority of the hosts in the tested region.