Intel Unveils Real-Time Deepfake Detector, Claims 96% Accuracy Rate
Intel technology has created the world’s first real-time deepfake detection software to root out this most-sophisticated of impersonation technologies.
The software uses tiny changes in image pixelation related to the movement of blood through human veins to detect whether a recorded video is real or fake with 96% accuracy.
Called FakeCatcher, it’s the first program that can catch deepfakes in real time, as most other detection apps require uploading videos for analysis, then waiting hours for results. They tend to need large data sets for deep learning programs to pour over.
Refreshingly, Ilke Demir, senior staff research scientist in Intel Labs opted for a more organic approach; one that also increases efficiency, as it’s capable of monitoring 72 video streams at once.
“FakeCatcher is the first approach that is telling us why we are real, why we are authentic,” said Demir. “It’s like a watermark of being human.”
Using databases of video camera footage or recorded speech, deepfakes use computer learning to construct audio or video recordings of real people saying and doing things they have never said or done. Politicians, radio personalities, or actors—people with long histories of recorded material, are at particularly elevated risk for deepfakes.
Businesses too, are at risk for these videos. Estimates range as high as $188 billion per year for American businesses to defend themselves against all manner of security threats originating from impersonations.
“FakeCatcher is a part of a bigger research team at Intel called Trusted Media, which is working on manipulated content detection—deepfakes—responsible generation and media provenance,” she said. “In the shorter term, detection is actually the solution to deepfakes — and we are developing many different detectors based on different authenticity clues, like gaze detection.”
Source: https://www.goodnewsnetwork.org/intel-unveils-real-time-deepfake-detector-claims-96-accuracy-rate/ |
