Scale-dependency in IP-based positioning of network clusters

Abstract

Although research in location-based services (LBS) is advancing well, the problem of obtaining a position for the user is still a major obstacle. Commonly available methods suffer from problems of availability, financial cost, and lack of precision or accuracy. The concept that IP-addresses tend to be spatially clustered which makes them attractive as a means for positioning. IP-based positioning would be applicable to any immobile device or interface, such as a computer or a wireless access point. Although it is believed that LBS equates to mobile computing, in reality the audience among static users in homes and offices may in fact be greater at this point in time. VRILS (varying resolution IP locating system) uses the relationship between network clusters and spatial clusters to provide positions for IP-addresses. It uses different levels of spatial precision to cope with conflicting locations within subnets, which enhances the chance of being able to provide a location. VRILS has been tested on the campus of Curtin University, where the positions of 461 IP-addresses were used in a network of over 20,000 computers. The outcome showed perfect results at the broadest spatial resolution of ‘campus’, and a reasonable result at the resolution of ‘building’. Randomness of IP-addresses across certain buildings was shown to strongly affect the accuracy. In general, it could be seen that with a relatively small amount of data, accurate positions could be obtained, but a lack of spatial clustering would decrease the efficiency to that of simple lookups.