Detecting a landing page that violates an online system policy based on a structural similarity between the landing page and a web page violating the policy

1. A method comprising:

receiving information describing a web page external to an online system, the web page violating a policy of the online system;

determining a hierarchical structure associated with the web page, the hierarchical structure comprising a first plurality of nodes representing elements of a first markup language document corresponding to the web page;

receiving a content item from a user of the online system, the content item comprising a link to a landing page external to the online system, where the link to the landing page is different than a link to the web page that violates the policy of the online system;

determining an additional hierarchical structure associated with the landing page, the additional hierarchical structure comprising a second plurality of nodes representing additional elements of a second markup language document corresponding to the landing page;

computing a measure of dissimilarity between the hierarchical structure and the additional hierarchical structure based at least in part on a difference between the first plurality of nodes and the second plurality of nodes; and

determining whether the received content item violates the policy based at least in part on the measure of dissimilarity.

2. The method of claim 1 , wherein each of the hierarchical structure and the additional hierarchical structure is a document object model (DOM) tree.

3. The method of claim 1 , wherein the measure of dissimilarity comprises a minimum edit distance between the hierarchical structure and the additional hierarchical structure, the minimum edit distance describing a minimum number of edits making at least a portion of the hierarchical structure identical to at least a portion of the additional hierarchical structure.

4. The method of claim 1 , wherein computing the measure of dissimilarity between the hierarchical structure and the additional hierarchical structure comprises:

generating a first hash representing at least a portion of the hierarchical structure and a second hash representing at least an additional portion of the additional hierarchical structure;

comparing the first hash and the second hash; and

computing the measure of dissimilarity based at least in part on the comparing.

5. The method of claim 4 , wherein generating the first hash and the second hash comprises:

truncating the hierarchical structure and the additional hierarchical structure;

generating a first set of characters representing a truncated portion of the hierarchical structure and a second set of characters representing an additional truncated portion of the additional hierarchical structure;

generating the first hash based on the first set of characters; and

generating the second hash based on the second set of characters.

6. The method of claim 5 , wherein comparing the first hash and the second hash comprises determining a Hamming distance between at least a portion of the first hash and at least a portion of the second hash.

7. The method of claim 6 , wherein computing the measure of dissimilarity based at least in part on the comparing comprises:

determining whether the Hamming distance is below a threshold number;

responsive to determining the Hamming distance is below the threshold number, computing a minimum edit distance between the hierarchical structure and the additional hierarchical structure; and

computing the measure of dissimilarity based at least in part on the minimum edit distance.

8. The method of claim 1 , wherein determining whether the received content item violates the policy is further based at least in part on a plurality of additional measures of dissimilarity between the additional hierarchical structure and a set of hierarchical structures associated with a set of additional web pages violating the policy.

9. The method of claim 1 , wherein determining whether the received content item violates the policy based at least in part on the measure of dissimilarity comprises:

providing the measure of dissimilarity between the hierarchical structure and the additional hierarchical structure as an input feature to a trained predictive model, the trained predictive model outputting the likelihood that the landing page violates the policy; and

receiving an output from the trained predictive model indicating the likelihood that the landing page violates the policy.

10. The method of claim 9 , wherein the trained predictive model is trained to predict the likelihood that the landing page violates the policy using a K-nearest neighbors algorithm.

11. The method of claim 1 , wherein determining whether the received content item violates the policy based at least in part on the measure of dissimilarity comprises:

predicting a likelihood that the content item violates the policy based at least in part on the measure of dissimilarity;

determining whether the predicted likelihood is at least a threshold likelihood; and

responsive to determining the predicted likelihood is at least the threshold likelihood, performing one or more corrective actions associated with the content item.

12. The method of claim 11 , wherein the one or more corrective actions are selected from the group consisting of: queuing the content item for manual review, disallowing presentation of the content item, communicating a warning to the user from whom the content item was received, and suspending an account on the online system associated with the user from whom the content item was received.

13. A computer program product comprising a non-transitory computer readable storage medium having instructions encoded thereon that, when executed by a processor, cause the processor to:

receive information describing a web page external to an online system, the web page violating a policy of the online system;

determine a hierarchical structure associated with the web page, the hierarchical structure comprising a first plurality of nodes representing elements of a first markup language document corresponding to the web page;

receive a content item from a user of the online system, the content item comprising a link to a landing page external to the online system, where the link to the landing page is different than a link to the web page that violates the policy of the online system;

determine an additional hierarchical structure associated with the landing page, the additional hierarchical structure comprising a second plurality of nodes representing additional elements of a second markup language document corresponding to the landing page;

compute a measure of dissimilarity between the hierarchical structure and the additional hierarchical structure based at least in part on a difference between the first plurality of nodes and the second plurality of nodes; and

determine whether the received content item violates the policy based at least in part on the measure of dissimilarity.

14. The computer program product of claim 13 , wherein the measure of dissimilarity comprises a minimum edit distance between the hierarchical structure and the additional hierarchical structure, the minimum edit distance describing a minimum number of edits making at least a portion of the hierarchical structure identical to at least a portion of the additional hierarchical structure.