The “deepfake challenge” to fight against misinformation

The "deepfake challenge" to fight against misinformation

Produce Deepfake to reduce deepfake?

What is Deepfake?

Deepfake, a contraction between “deep learning” and “fake”, is a video produced or altered by artificial intelligence. This practice of Hyperfaking or intelligent permutation of faces is indeed an image synthesis technique based on artificial intelligence used mainly to superimpose existing images and videos on other images and/or videos.


Deepfake: a problematic and dangerous phenomenon that compromises the news


This phenomenon of fallacious content, made highly credible by artificial intelligence, is now more and more common and widespread on social networks. Deepfake techniques make personalities say things they have never said and therefore pose a major challenge to the legitimacy of online information.

In addition, thanks to the development of AI, anyone can create a Deepfake relatively easily and without any special technical skills by downloading simple software.

  • Social networks then contain more and more fake content and everyone can then produce such fake content in order to serve their interests, manipulating the opinions of other users.

With these faked videos, increasingly convincing thanks to the development of AI algorithms, and considered real by nearly a third of untrained human subjects, the misinformation becomes real and expands.

  • These false contents seem increasingly real and convey false information in which many users entirely believe.

This phenomenon is all the more problematic in the context of the upcoming US presidential elections, which will be held in November 2020.

It then becomes urgent to act, but how to fight against this phenomenon of fake news which is becoming more and more credible and widespread?


Fight Deepfakes by making them more easily detectable through the development of AI tools.


Facebook has decided to take matters into their own hands and fight this issue.

For this reason, the social network has decided, in partnership with Microsoft and several prestigious universities (MIT and Oxford), to put 10 million euros at stake in a competition to find the best possible anti-deepfake tool: that is the Deepfake Challenge.

The objective of the Challenge being to fight against misinformation and fake news, by making it easier to detect fake videos through the development of artificial intelligence tools, will restore a climate of trust on social platforms relaying this type of content.


Produce Deepfake to reduce deepfake?


The social network explains that current identification tools are based on the detection of classic artifacts (inconsistencies in shadows, approximate detours) or poorly synchronized movements, but are not powerful enough.

In order to meet this challenge and to be able to properly develop the AI detection tools, Facebook then produced Deepfake itself which it makes available to challengers, participants in the challenge.

To encourage participation, the social network will allocate more than $10 million for this purpose, some of it in the form of prizes. The company headed by Mark Zuckerberg has announced that it will participate in the competition but will not accept any compensation.

The challenge will begin at the end of the year, when sufficient data will be available for participants, and will end in March 2020.


Let’s hope that thanks to this Facebook initiative, the era of deepfake and photo manipulation will soon be over.

In the meantime, stay informed about the evolution of the Deepfake Challenge with Eminence!