Intel Unveils 96% Effective Defeat Detection Tool
Intel’s detector is able to recognize manipulated images, videos and audio sequences in
milliseconds with 96% accuracya first according to what can be read in a blog post from the company.
What is a deep fake?
The deep fake, called fake deep in English, is an altered image, video or audio sequence in which artificial intelligence (AI) is used to replace, for example, the face of a political figure with another and make him say things that he never pronounced.
According to Intel, the majority of such tools examine raw image data for signs of inauthenticity. It is also often necessary to
download the videos for analysis, then wait hours for the resultsemphasizes the company.
Detect blood flow
Instead, FakeCatcher looks for clues of authenticity in
what makes us human: subtle blood flow in video pixels.
When our heart pumps blood, our veins change color. These blood flow signals are collected from across the face, and algorithms translate these signals into spatiotemporal maps. Then, thanks to deep learning, we can instantly detect if a video is real or fake.we detail in the Intel blog post.
A growing threat
Tampered content is a growing threat, according to the company.
Deception due to deep fakes can cause harm and lead to negative consequences, such as reduced trust in the mediabelieves Intel.
FakeCatcher, designed in partnership with the State University of New York at Binghamton, could thus be useful for the media by allowing them to authenticate content circulating online. Social networks could also benefit from it by flushing out harmful hyper-deceptions before they circulate on its platforms.