While Deepfakes can be fun, they have increasingly been used to spread fake news and fake pictures of people, especially celebrities. Over time, spotting one and getting harder but Intel has a solution to sort this out. The company has introduced the FakeCatcher technology, which can detect a fake Deepfake in real time. Check out the details below.
Certified Human: How New Intel Technology Detects Deepfakes in Real Time | Intel
FakeCatcher can call out a fake video with 96% accuracy and is the world's first to show results in milliseconds. It was developed by Ilke Demir, a senior researcher at Intel Labs in collaboration with Umur Ciftci from the State University of New York.
This AI-based Deepfake detection tool tries to look for real parts of a video instead of looking for fake ones, much like most detectors out there. So what is it looking for? FakeCatcher assesses the blood flow of people in the pixels of the video, which is considered the most human.
Now, how does this happen? FakeCatcher is based on photoplethysmography (PPG), which uses a light source to detect variations in blood circulation. The veins change color as the heart pumps blood and this is recorded as PPG. The PPG signals are taken from the 32 locations on the face and then translated into spatiotemporal maps using Intel algorithms. This is when it can be judged whether a video is fake or real.
