A system and method for determining similarity of nodes on a graph using product similarity scores is described. A graph of nodes can be broken into a plurality of bi-partite graphs. For each bi-partite graph, the probability of getting from a first node to a second node through random walk can be determined. This probability can be multiplied with the probability of getting from a second node to a first node through a random walk to determine a product score for the two nodes. In some instances, the product score can be normalized using normalized conditional probabilities. These normalized probabilities can be used to determine the similarity between two nodes, e.g., entities, on a graph. This can be used to detect related sites, determine fraudulent sites, and increase security.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.