American multinational corporation and technology company Intel has claimed to introduce the world's first mile-time deep-fake detector. The fake catcher is said to have an accuracy rate of 96 percent and helps in video identification by analyzing blood flow in video pixels using advanced photopathography (PPG). . According to a report, Apke Damier, senior staff research scientist at Intel Labs, developed the fake catcher detector in collaboration with Amur Safety of the State University of New York at Benjaminton. The real-time detector uses Intel hardware and software and runs on a server and interfaces through a web-based platform. FakeCatcher differs from deep learning-based detectors in that it looks for authentic signals in real videos instead of looking at raw data to look for signs of inauthenticity. The way it works is based on PPG, a method that measures the amount of light that is absorbed or reflected by blood vessels in living tissue. As the heart pumps blood, the veins change color and these signals are monitored by technology to determine if a video is fake. Speaking to the online news platform VentureBeat about the technology, AK Demir said that the fake catcher is unique because PPG signals have not been applied to the deepfake problem before. The detector collects these signals from 32 locations on the face before deciding whether the video is real or fake. Algorithm then transforms them into spatiotemporal maps. Deepfake videos are a growing threat worldwide. Companies will spend up to $188 billion on cybersecurity to combat them, according to technology research and consulting firm Gar Tanner. Currently, deepfake detection apps require videos to be uploaded for analysis and results can typically take hours. Intel says social media platforms can take advantage of the detector to prevent users from uploading malicious deepfakes.
 |
Deepfakes have targeted prominent politicians and celebrities. In a viral TikTok video last month, US President Joe Biden appeared to be singing the children's song Baby Shark instead of the national anthem. Efforts to detect DeepFax have also been hampered by ethnic bias in the datasets used to train it. According to a 2021 study from the University of Southern California, some detectors showed a 10.7 percent difference in error rate by racial group.