A View into YouTube View Fraud

Abstract

Social media platforms are driven by user engagement metrics. Unfortunately, such metrics are susceptible to manipulation and expose the platforms to abuse. Video view fraud is a unique class of fake engagement abuse on video-sharing platforms, such as YouTube, where the view count of videos is artificially inflated. There exists limited research on such abuse, and prior work focused on automated or bot-driven approaches. In this paper, we explore organic or human-driven approaches to view fraud, conducting a case study on a long-running YouTube view fraud campaign operated on a popular free video streaming service, 123Movies. Before 123Movies users are allowed to access a stream on the service, they must watch an unsolicited YouTube video displayed as a pre-roll advertisement. Due to 123Movies' popularity, this activity drives large-scale YouTube view fraud. In this study, we reverse-engineer how 123Movies distributes these YouTube videos as pre-roll advertisements, and track the YouTube videos involved over a 9-month period. For a subset of these videos, we monitor their view counts and metrics for their respective YouTube channels over the same period. Our analysis reveals the characteristics of YouTube channels and videos participating in this view fraud, as well as the efficacy of such view fraud efforts. Ultimately, our study provides empirical grounding on organic YouTube view fraud.