A workaround to indirectly estimate the distribution of classes for an Autonomous System

Prefix-classes group network prefixes that are equivalent with respect to the choice of the next-hop border gateway within an Autonomous System (AS). Clustering the routing information is such way significantly reduces the volume of data needed to analyze or design a network, from several million incoming updates to a few hundred classes, while still finding optimal routes. Existing methods for calculating prefix classes rely upon detailed internal data of the AS under study, which is rarely available. This research proposes a new stochastic method to estimate prefix classes for any AS using only publicly available information. This probabilistic model was tested on a real network and achieved a high degree of accuracy.