Intel released what it claims is the world’s first deepfake detector on November 14, which can return results in milliseconds with a 96% accuracy rate. Namely, FakeCatcher, which utilizes photoplethysmography to measure the blood flow of a human subject in video pixels (PPG).
“The majority of deep learning-based detectors examine unprocessed data to discover signals of inauthenticity and determine what is wrong with a video. Intel revealed in a blog post that FakeCatcher looks for authentic clues in real films by analyzing what makes us human – subtle “blood flow” in the pixels of a video.
Deepfakes are typically fake videos that employ artificial intelligence to duplicate a person’s face onto another in an attempt to trick individuals into believing that they are the subject of the clip. In deepfakes, blood flow may be nonexistent, in contrast to actual videos.
“When our hearts pump blood, the color of our veins changes. These blood flow signals are gathered from the entire face, and algorithms convert them into spatiotemporal maps. Then, via deep learning, we can instantaneously determine if a video is authentic or false,” Intel stated.
According to Intel, prospective use cases for the technology include social media platforms preventing its users from submitting damaging deepfake films, global news organizations avoiding boosting altered content, and non-profit groups “democratizing detection of deepfakes for everyone.”
Intel reasoned that deepfake films are a growing concern and referenced Gartner’s estimate that businesses spend as much as USD 188 billion on cybercrime solutions. FakeCatcher’s ability to recognize whether a video is false or not in real-time inhibits the immediate dissemination of potentially influential deepfake. Other existing deepfake detection tools typically need the uploading of films for examination, resulting in hours of waiting for results.
“Deception caused by deepfakes can cause harm and have undesirable implications, such as a loss of faith in the media.” Intel stated that FakeCatcher helps reestablish trust by allowing consumers to discriminate between legitimate and false content.