Identifying fraudulent requests for content

A computing device comprising:

• a processor;
• and
• memory comprising processor-executable instructions that when executed by the processor cause performance of operations, the operations comprising:
• receiving a request for content from a first device;
• identifying a first user profile associated with the first device, wherein the first user profile comprises at least one of:
• activity information associated with the first device;
• demographic information associated with the first device;
• or
• interest information associated with the first device;
• determining, based upon the request for content, a resource associated with the request for content;
• analyzing a user profile database based upon the first user profile to identify a set of user profiles similar to the first user profile;
• generating a relevance score corresponding to a probability that a first user with the first user profile accesses the resource and associated with the request for content, the generating the relevance score based upon (i) the resource determined based upon the request for content, (ii) the set of user profiles similar to the first user profile and (iii) the first user profile associated with the first device from which the request for content was received;
• determining whether the request for content is fraudulent by comparing:
• the relevance score corresponding to the probability that the first user with the first user profile accesses the resource, associated with the request for content and generated based upon (i) the resource determined based upon the request for content, (ii) the set of user profiles similar to the first user profile and (iii) the first user profile associated with the first device from which the request for content was received, with a threshold relevance;
• and
• responsive to a determination that the request for content is fraudulent, at least one of:
• not transmitting a content item, associated with the request for content, to the first device;
• or
• discarding the request for content,
• wherein at least one of:
• the first user profile indicates one or more languages associated with the first device;
• or
• the generating the relevance score comprises:
• determining, based upon the set of user profiles and the first user profile, an expected behavior associated with the first user profile using one or more machine learning techniques;
• and
• determining a difference between the expected behavior and a second behavior associated with the resource, wherein the relevance score is indicative of the difference.

A non-transitory machine readable medium having stored thereon processor-executable instructions that when executed cause performance of operations, the operations comprising:

• receiving a request for content from a server associated with a resource, wherein the request for content is associated with a first device;
• identifying a first user profile associated with the first device, wherein the first user profile comprises at least one of:
• activity information associated with the first device;
• demographic information associated with the first device;
• or
• interest information associated with the first device;
• analyzing a user profile database based upon the first user profile to identify a set of user profiles similar to the first user profile;
• generating a relevance score corresponding to a probability that a first user with the first user profile accesses the resource and associated with the request for content, the generating the relevance score based upon (i) the resource associated with the request for content, (ii) the set of user profiles similar to the first user profile and (iii) the first user profile associated with the first device associated with the request for content;
• determining whether the request for content is fraudulent by comparing:
• the relevance score corresponding to the probability that the first user with the first user profile accesses the resource, associated with the request for content and generated based upon (i) the resource associated with the request for content, (ii) the set of user profiles similar to the first user profile and (iii) the first user profile associated with the first device associated with the request for content, with a threshold relevance;
• and
• responsive to a determination that the request for content is fraudulent, at least one of:
• not transmitting a content item, associated with the request for content, to the first device;
• or
• discarding the request for content,
• wherein at least one of:
• the first user profile indicates one or more languages associated with the first device;
• or
• the generating the relevance score comprises:
• determining, based upon the set of user profiles and the first user profile, an expected behavior associated with the first user profile using one or more machine learning techniques;
• and
• determining a difference between the expected behavior and a second behavior associated with the resource, wherein the relevance score is indicative of the difference.