Facebook can now detect who created deepfakes

Deepfake new technology makes adults tremble. Nowadays, thanks to public applications, it is possible to fool even the most careful person by replacing one person’s face with another. The consequences and risks of this are many and serious to say the least; Therefore, large companies and scientists are now looking for ways to overcome this situation. to avoid the worst. Facebook set itself this mission and partnered with Michigan State University to achieve it.

The result of this collaboration could not be more remarkable. In fact, the artificial intelligence developed by researchers is not satisfied with this. Detecting the existence of deepfake in a video. He is talented too find trace of its creator. Specifically, it can tell which model was used to create the hoax and trace it back to the source of the broadcast with the correct information in hand.

Facebook can tell who created deepfake

Technology “It is based on discovering unique patterns behind the AI ​​model used to create a single deepfake.”Explain to researchers. “We start with image correlation and then try to explore the properties of the model used to create the image. […] By generalizing image correlation to database detection, we can gain more insight into the generative model used to create deepfakes ». This means that the artificial intelligence is analyzing a deepfake that uses a model it does not know, and may not be able to detect its existence.

Despite this flaw, the technology has many qualities. In particular, it can: detect multiple deepfakes belonging to the same model category and thus we conclude that they come from one and the same source. To do this, it analyzes the “fingerprint” of the video. Just like photos taken with different devices and structured in various ways, deepfakes are created according to their own parameter systems. It is the latter that makes them unique and therefore traceable by definition.

Source : Facebook