Using machine learning techniques to detect imposter pages in an online system

摘要

An online system detects imposter pages based on machine learning techniques. The online system maintains a plurality of authenticated pages and a plurality of unauthenticated pages, each of which is associated with a name and an image. From the plurality of unauthenticated pages, the online system filters out one or more unauthenticated pages that are associated with names of authenticated pages to obtain a group of candidate pages. Further, the online system pairs each candidate page up with an authenticated page. The candidate page has a name and/or image similar to the authenticated page. The online system inputs the candidate page and the authenticated page into a trained model. The trained model outputs an imposter score indicating a likelihood that the candidate page is an imposter page. The online system takes actions on the candidate page based on the imposter score.