Oxford University to create deep fake videos in bid to beat political manipulation on Facebook

tmg.video.placeholder.alt reAjA_bQGig

Oxford University scientists will create "deepfake" videos to test the accuracy of Facebook’s detection technology in the fight against political manipulation online.

Increasingly technology is allowing for more and more realistic “deepfakes” to be created, using deep learning or artificial intelligence to make fake videos or audio that are almost indiscernible from the real thing. 

Oxford University will contribute to a new data set of realistic deepfake videos and help Facebook to train its algorithms to detect true videos from manipulated ones.

The university is the only UK institution among the cohort of universities, which includes MIT and the University of Berkeley, helping the social media giant do this work.

Professor Philip Torr said: "Manipulated media being put out on the internet, to create bogus conspiracy theories and to manipulate people for political gain, is becoming an issue of global importance, as it is a fundamental threat to democracy and freedom. 

“I believe we urgently need new tools to detect and characterize this misinformation, so I am happy to be part of an initiative that seeks to mobilize the research community around these goals — both to preserve the truth whilst pushing the frontiers of science.”

Facebook, which has dedicated over $10m (£8.1m) to this project called the "Deepfake Detection Challenge", is under pressure to stop fake videos of high profile politicians from spreading false information ahead of major elections, including the US presidential vote in 2020. 

tmg.video.placeholder.alt dDGcWDuZ2bA

Last year US lawmakers warned that such faked videos could be damaging to national security.  

The results of this challenge will be shared with the rest of the industry, Facebook said, and will be overseen by a cross-sector coalition of organizations including Facebook, WITNESS, Microsoft, and others in civil society and the technology, media, and academic communities.

Facebook confirmed that no user data will be used in this trial.

Earlier this year, scientists were able to create a convincing deep fake video of the Mona Lisa into a real life person based solely on the image created by Leonardo Da Vinci, ramping up concerns among experts that deepfake videos will only need a still image to create a false video.

Chinese app Zao, which went viral earlier this month, allowed users to swap their faces onto celebrities' bodies and insert themselves into television series and movies and create seamless deepfakes for fun. 

License this content