Neighborhood clustering for web spam detection

1. A web spam detection system comprising:

a server including at least one processor and at least one computer-readable storage media coupled to the processor, wherein the computer-readable storage media includes at least the following components:

a propagation component to assign initial spam and non-spam scores to one or more nodes, wherein the initial spam and the non-spam scores are output from supervised or unsupervised learning; and

a processing component that iteratively applies local effects to the one or more nodes and updates the initial spam and the non-spam scores for the one or more nodes in order to facilitate web spam detection.

2. The system of claim 1 , further comprising a component to determine if a node links to many nodes with high spam scores, where high spam scores are determined via a threshold.

3. The system of claim 2 , further comprising a component to increase or decrease a spam score associated with the one or more nodes.

4. The system of claim 3 , further comprising a component to determine if incoming links of a node originate from nodes with a low page rank, where low page rank is determined from a threshold.

5. The system of claim 1 , wherein the propagation component iteratively distributes spam scores throughout a graph to provide more accurate spam scores for unlabeled nodes.

6. The system of claim 5 , wherein the graph is associated with a domain graph, a weighted domain graph, a diluted domain graph, or a full web graph.

7. The system of claim 1 , further comprising a component to detect one or more link structures of a graph.

8. The system of claim 7 , wherein the component determines how many spam pages are linked to other spam pages in order to increase a ranking value to a search engine.

9. The system of claim 1 , further comprising a component to determine if many domains with a low spam score link to a domain, where a low spam score is determined via a threshold value.

10. The system of claim 9 , further comprising another component to decrease the low spam score of the domain.

11. The system of claim 1 , further comprising a component to determine if a domain links to many domains with a high spam score, where the high spam score is determined via a threshold value.

12. The system of claim 11 , further comprising another component to decrease the high spam score of the domain.

13. The system of claim 1 , further comprising assigning one or more thresholds to determine spam scores for a domain.

14. The system of claim 13 , further comprising two thresholds for the determined spam scores that are assigned as “very low” and “high”, and two thresholds which determine a desired number of domains.

15. The system of claim 14 , wherein a desired number of domains is determined from an absolute and a relative number of in-coming or out-going links of very low or high spam scores, respectively.

16. The system of claim 14 , further comprising one or more non-linear functions that replace the thresholds.

17. A web spam detection method, comprising:

generating initial spam and non-spam scores for one or more nodes of a graph, wherein the initial spam and the non-spam scores are output from supervised or unsupervised learning;

applying the initial spam and the non-spam scores to the nodes based on a link structure; and

updating the initial spam and the non-spam scores for the one or more nodes on an iterative basis in order to facilitate web spam detection.

18. The method of claim 17 , further comprising determining if a node links to a number of nodes with high spam scores, where the high spam scores and the number of nodes are determined via a threshold.

19. The method of claim 17 , further comprising a component to increase or decrease a spam score associated with the one or more nodes.

20. A web spam detection system, comprising:

a processor;

means operated by the processor for determining initial spam and non-spam scores for one or more nodes of a graph based in part on a link structure, wherein the initial spam and the non-spam scores are output from supervised or unsupervised learning; and

means operated by the processor for updating the spam and non-spam scores for the one or more nodes in order to facilitate web spam detection.