
Researchers at Intel have developed a real time tool to detect fake videos with 96% accuracy.
FakeCatcher builds on open source computer vision and AI software technologies to analyse blood flow signals from all over the face in the video. Algorithms translate these signals into spatiotemporal maps to assess whether a video is real or fake.
“Deepfake videos are everywhere now. You have probably already seen them; videos of celebrities doing or saying things they never actually did,” said Ilke Demir, senior staff research scientist in Intel Labs
Demir developed the tool with Umur Ciftci from the State University of New York at Binghamton. Using Intel hardware and software, it runs on a server and interfaces through a web-based platform.
The OpenVino framework is used to run AI models for face and landmark detection algorithms. Computer vision blocks were optimized with a a multi-threaded software library called Integrated Performance Primitives and the OpenCV toolkit for processing real-time images and videos. Inference blocks were optimized with Intel’s Deep Learning Boost tool and Advanced Vector Extensions 512, and media blocks were optimized with Advanced Vector Extensions 2.
The researchers also used the Open Visual Cloud project to provide an integrated software stack for the Intel Xeon processors, running up to 72 different detection streams simultaneously.
There are several potential use cases for FakeCatcher. Social media platforms could use the technology to prevent users from uploading harmful deepfake videos. Global news organizations could use the detector to avoid inadvertently amplifying manipulated videos. And nonprofit organizations could employ the platform to democratize detection of deepfakes for everyone.
