In an effort to boost research and development in the context of deepfake detection, Google has shared a database containing 3,000 deepfake videos with the new FaceForensics benchmark, a research project by researchers at the Technical University of Munich and the University Federico II of Naples. Deepfakes are audio or visual content doctored by artificial intelligence (AI). They are considered a major threat because they allow threat actors to spread disinformation and influence public opinion by making it seem like influential individuals including government, corporate and military leaders, candidates in democratic elections, scientists and celebrities, said or did things they didn’t actually say or do.
The videos released by Google were produced with the help of paid actors. After recording hundreds of videos, Google researchers used publicly available tools to create thousands of deepfake videos. Researchers at the FaceForensics benchmark project can use the combination of legitimate and deepfake videos to test and improve algorithms designed to spot manipulated footage.
Read more: Google’s war on deepfakes: As election looms, it shares ton of AI-faked videos