Google has released a trove of deepfake videos as part of an effort to combat the spread of manipulated videos that can be used to harm people.

Deepfakes

The so-called deepfakes are manipulated video and audio clips, which first emerged in late 2017. The internet has since witnessed an increase in synthesized media clips.

Many of these manipulated clips may be intended to be humorous but some pose harm to individuals and even the society.

A report from the New York University Stern Center for Business and Human Rights released earlier this month, for instance, warned that bad actors can use altered videos of candidates to manipulate the 2020 U.S. presidential election.

Earlier this year, a digitally altered video showing U.S. House of Representatives speaker Nancy Pelosi slurring drunkenly through a speech gained widespread attention on social media, highlighting the potential danger of deepfakes when used as a tool to target celebrities and politicians.

Google's Contribution To Deepfake Detection Research

In a bid to combat the proliferation of these altered videos, Google released a dataset that contains 3,000 deepfake videos that it created.

In a blog post on Tuesday, Nick Dufour of Google Research and Andrew Gully of Jigsaw wrote that Google worked with paid and consenting actors to record hundreds of videos, which were then used to generate thousands of deepfakes using publicly available deepfake generation methods.

Google is releasing the fake videos and the real ones to directly support deepfake detection efforts.

The tech company hopes that releasing the manipulated videos as part of the FaceForensics benchmark can help researchers and other experts come up with new ways to identify and combat these videos. It could also provide for tools that can help the public and publishers identify deepfakes.

"We firmly believe in supporting a thriving research community around mitigating potential harms from misuses of synthetic media, and today's release of our deepfake dataset in the FaceForensics benchmark is an important step in that direction," the blog post read.

The dataset is available for free to the research community for use in developing synthetic video detection methods. Google also said that it will be updating the dataset with new videos as the deepfake technology continues to evolve.