We all hear about the scandal of Zague. It is not necessary to remember unnecessary details, but when the former footballer and analyst on sports attempt to explain the embarrassing moment via Twitter, attributed to hackers and the “photoshop of video”. The reference driver is the Deepfake, the ability to alter faces in video using Artificial Intelligence to forge identities.
— Zague (@LRZague) June 12, 2018
The Deepfake is giving a lot to talk about in the digital world and media. Their sophistication has resulted in videos with characters false emulating public figures to become viral, with thousands, but millions of users that’s a given.
How to recognize a Deepfake?
That is why scientists at Cornell University in New York, united States of America, are spreading methods and the key to recognizing this type of fraud in video and avoid that continue to commit abuses and bullying digital against innocent people.
According to the team confirmed by the specialists Yuezun Li, Ming-Ching Chang and Siwei Lyu, the key is in the eyes, eye movements and blinks. “Our method is based on the detection of the eye blinks in the videos, which is a physiological sign that is not well represented in the fake videos,” explain the researchers.
So the key to knowing if a video is a deepfake is to look in the eyes of the alleged character in question: The protagonists of these clips are fake they don’t blink like humans, they do, on average, 17 times per minute. If a person does not blink in a video, it is very likely that it is apocryphal. “The algorithms of the ‘deepfake’ inherit this basic defect,” explains Siwei Lyu.
There are other signs to look for to detect deepfakes, these are several which lists the BBC:
- The source that published it is unknown or unreliable: for example, a string of WhatsApp. Analyzes the source.
- The places and dates do not match: pay attention to these and other details, like the logo or the place from where it was recorded.
- Pay attention to the mouth: besides the eyes, the mouth movements can be revealing.
- Pay attention to the message: it’s usually appealing to emotion and it seeks to generate certain reactions in the listener.
- Watch what happens in slow motion: slows down and for parts of the video to look more closely at the transitions.
The future of the deepfake, much more sophisticated
Although these keys and up to technology created by the same team of Cornell help to monitor the eye movements of the protagonists in the videos and with an accuracy of 99%, the specialists estimate that the deepfake to improve much in the short and medium and will be extremely difficult to recognize.
“Probably nothing that we can do more that try to fight the deceptions when they come to light, putting pressure on social networking companies to fight decidedly against the misinformation and rely on our eyes a little less each day,” wrote Kevin Roose, a journalist for the source of technology in The New York Times, about it.