A deepfake detector was trained for eight hours to spot false videos of Volodymyr Zelensky.
Hany Farid at the University of California, Berkeley, and Matyas Bohacek of the Johannes Kepler Gymnasium in the Czech Republic developed detection capabilities for faces, voices, hand gestures and upper body movements, and it had 100% accuracy in spotting three fake videos of the Ukrainian president, which had altered his mouth movements and spoken words.
Farid told New Scientist magazine: “It can say, ‘Ah what we observed is that with President Zelensky, when he lifts his left hand, his right eyebrow goes up and we are not seeing that.’
“We always imagine there’s going to be humans in the loop, whether those are reporters or analysts at the National Security Agency, who have been able to look at this being like, ‘Why does it think it’s fake?’ “
But the time-consuming training process means it is less suitable for identifying deepfakes involving members of the public or non-consensual videos of sexual acts.
However, the researchers have built another detector to spot false videos of US leader Joe Biden and are considering similar models for the likes of Vladimir Putin, Elon Musk and Xi Jinping of China.
They plan to make it available to certain government and news organisations.
