DeepFakes: uncovering hardcore open source on GitHub

ABSTRACT

In November 2017, a subreddit dedicated to ‘DeepFakes’ brought the work of machine-learning video face-swapping to widespread attention: the new ease of pornographic fakeswapping (focused primarily on celebrities, but also used more broadly to generate a type of revenge porn) became a content regulation topic of discussion on sites such as Pornhub and Reddit. Although the spaces dedicated to sharing these videos were gradually banned or moderated in early 2018, the essential project remains: a series of codebases shared on GitHub, actively situated for open source contributions. In this case study, we will examine the role of GitHub as a platform in the spread of DeepFakes, and interrogate the intersections of the open source ethos and Reddit's approach to moderation, ethics, and anonymity. Ultimately, we find that the DeepFakes project has intentionally spread to spaces with limited moderation, which has allowed the continued creation of non-consensual pornography.

KEYWORDS:

• Open source
• DeepFake, social media
• involuntary pornography
• ethics
• anonymity
• moderation

Disclosure statement

No potential conflict of interest was reported by the authors.