Everybody has already been confronted once with “deepfake”. Those videos, manipulated with the help of Artificial Intelligence are becoming more and more realistic. The physical imitation is effective so is the voice reproduction. 

In order to counter this threat, tools, also based on artificial intelligence, in particular Machine Learning, are developing to spot distortions. Deepfake is in a full expansion and experts all over the World plan more resources to dedicate their research to respond to it.


Fake videos seem not as much realistic to fool everyone. Experts confirm that the quality of deepfake depends on creator’s skills; the face swap in a video brings imperfections that any Human-being can detect. It will take years before a skillful eye can recognise a deepfake but for the moment, only machine learning is able to do so. 


In the meantime, Microsoft is working on its own deepfake detection tool as well as social media, which are taking precautionary measures, especially because those videos are relying on them. 


For example, In March 2020, Twitter introduced a new rules against sharing manipulated media contents which can cause damages. About Facebook and Google, they both shared fake pictures database to help Researchers to develop detection skills. 



Publié le 17 mars 2020