Intel Corporation has created the first real-time detector of deepfakes (fake images or videos that appear to be real, thanks to technology). In their announcement, the company introduced it under the name ‘Fakecatcher’ and they promised Accuracy rate of 96%.
According to Intel, your system Analyzes “blood flow” in pixels To know whether the content of the video is real or not in milliseconds. To develop this technology, the company uses its own hardware and software that run on a server and interact via a web-based platform.
Ilke Demir, principal research scientist on the staff of Intel Labs, was the creator of the system and collaborated with Omar Sifti of the State University of New York (Binghamton). experts focus on Photoplethysmography (PPG)A technique that measures the amount of light absorbed or reflected by blood vessels in living tissue.
The color change that occurs when blood moves through the veins is undetectable to the human eye, “but it is computational visualization“, Demir explains to VentureBeat. The PPG indicates that FakeCatcher detects that there are 32 places on the face.
Demir explains that with such signals, maps are created: “We trained a convolutional neural network on PPG maps to classify them as fake and real.” Then, they use technologies like the Deep Learning Boost Framework or Intel’s Advanced Vector Extensions 512 to make it run in real time and hold “up to 72 simultaneous detection streams.”
As deepfakes become more realistic, these detectors are becoming more important. fake pictures and videos They are often used to commit cybercrime And it is becoming very difficult to detect it with the naked eye.
Sign up for our newsletter and get the latest technology news delivered to your inbox.