FakeCatcher, Intel’s new deepfake detection platform, claims 96 percent accuracy in real-time deepfake detection. And the secret is our blood.
Within system using blood flow to detect deepfake video – BBC News
A new deepfake detector that claims to deliver near-instant, highly accurate results has been unveiled by Intel, using what the company calls a far more reliable method than reverse engineering. What FakeCatcher does differently, which Intel believes makes it more unique, is look at the video itself rather than the raw data behind it.
This new approach analyzes potential deepfakes in real time, using various tools such as facial detection AI models and algorithms. It also looks for (and can reportedly detect) “blood flow,” through subtle changes in pixels, which occur as our blood naturally flows. These pulses are picked up by FakeCatcher, translated into visible maps, and then checked for inconsistencies, all in a very short time.
Intel believes real-time analysis will make a big difference, as other deepfake detectors can take hours to process and often require uploading suspicious media first.
