As part of its AI efforts, Intel has produced FakeCatcher, a technology that can detect fake videos with 96% accuracy. Intel’s deepfake detection platform serves as the world’s first real-time deepfake detector with results in milliseconds.

“Deepfake videos are everywhere now. You’ve probably seen them already; videos of celebrities doing or saying things they’ve never actually done” said İlke Demir, a senior staff research scientist at FakeCatcher and Intel Labs.

Most deep-learning-based detectors look at raw data to try to find non-authentic signs and determine what’s wrong with a video. In turn, FakeCatcher looks for unique clues in real videos by assessing the fine blood flow in the video’s pixels. When blood is pumped, blood flow signals are collected all over the face, with changing vessel colour. Algorithms transform these signals into spatial-temporal maps. FakeCatcher then uses deep learning to instantly detect whether a video is real or fake.

Deepfake videos have become a growing threat. Scams caused by deepfake videos can harm users and lead to negative consequences such as a decrease in the user’s trust in the media. The FakeCatcher system, developed to eliminate the threat that causes such problems, aims to prevent many problems.