The quality and speed at which videos can now be faked using neural networks and deep learning sue hope to make the upcoming presidential election even more of a nightmare . But by tap somethingoverlooked in current deepfake technique , researchers have rule an automatize manner to spot bogus video recording .
Deepfake television are far from staring right now . Created from giant libraries of image scrap from the internet , they ’re often generated at scurvy resolution ( which helps hide imperfections ) and appear overly compressed . But the technology is better at a startling rate , and flaws in the summons , like deepfake picture that wereeasy to spot because the topic never blinked , are quickly improved to make them more and more believable .
https://gizmodo.com/most-deepfake-videos-have-one-glaring-flaw-1826869949
It ’s an weapon system slipstream that neither side is going to stomach down from any time soon , but researchers from UC Berkeley and the University of Southern Californiabelieve they ’ve developed the next weaponwhen it come to battling , or only just accurately identifying fake videos . Using a like process to how deepfakes are created — by studying existing footage of the current crop of presidential hopefuls — they civilize an AI to count for the presence of each person ’s “ soft biometric ” signature . It sounds complicated , but when speaking , we all have elusive but unique way we move our bodies , head , hand , eyes , and even lips . It ’s all done subconsciously — you do n’t realize your body is doing it , nor does your mind immediately recognize when someone else is — but as a event , it ’s a contingent that current deepfake processing techniques do n’t take into account when creating a role player .
In testing , the new AI was capable to accurately spot deepfakes at least 92 pct of the metre , including telecasting created using several technique , and those with degraded mental image quality due to TV files being overly press . The researcher plan to further meliorate the AI ’s success rate by taking into invoice the singular measure and characteristics of a soul ’s actual voice . But the realness is that deepfake techniques are germinate and improving at such a rate that they ’ll likely compensate and be able to fool this AI before 2020 even make it . This research represents a battle get ahead , but the war for truth online is still going to wage on .
Deepfakesuc berkeley
Daily Newsletter
Get the best technical school , skill , and cultivation word in your inbox day by day .
News from the future , delivered to your present .
You May Also Like
