A First Look at Scams on YouTube
Elijah Bouma-Sims, Brad Reaves · 2021
YouTube channel of the scam video owner.Finally, we revisit these videos five months later to determine liveness.In this paper, we use the collected dataset to address the following research questions: RQ1 How does the metadata differ between scam and non-scam videos?Scam videos are younger than non-scam videos, have fewer views and less comment engagement, and are posted by accounts with less activity than non-scam videos.RQ2 How do scammers monetize scams on YouTube?The vast majority of scammers redirected to external websites, many of which use "Cost per action" (CPA) monetization tools like surveys to make money.Further, most scams in our dataset were on search terms related to gift cards or mobile games.RQ3 Can a classifier use metadata alone to distinguish scam and non-scam videos?Statistical fields related to channel size and popularity provided the most mutual information.Additionally, the presence of gift card and mobile game-related words in metadata was found to be significant in discriminating between videos.