They discover a simple way to detect deepfakes during a video call

When talking about deepfake, reference is made to a technique, powered by artificial intelligencewhich is applied with the purpose of creating videos replicating the face of a person in the face of another, thus managing to imitate their facial expressions and adjusting them to the movement of the body as if it were the person in question.

While most deepfakes are easily recognizable, there are some that put our eyes to the test.

In the field of video calls, there are filters made from deepfakes that are quite credible. Within a communication made in low or medium resolution, make a deep fake in these circumstances requires powerful enough software and hardware to turn one person into another without difficulty.

In that sense, it is a problem to try detect a deepfake in real timewhich has meant that malicious people can take advantage of this situation to carry out pranks or scams, by adopting the identity of recognized people in the world of politics and business.

In this regard, Mikel Anderson published an article in the journal Metaphysic where he mentioned the existence of a low-tech procedure with which it is possible to detect deepfakes during a video call.

The secret is to put yourself in profile from time to time

All you have to do is ask the person put your face in profile. The reason is that the algorithms used for the creation of the deepfakes they don’t have too many records photographs of the person at this anglebeing that these are more effective when registering faces that look completely straight ahead, since in this position the facial features can be appreciated with more detail.

At the moment that the person with the deepfake places their face in profile, it is distorted, revealing parts of the original face.

In this article you can also find information about how these algorithms work, as well as the different actions that are carried out to recreate faces.

Other points discussed in the article mention the source from which the data is taken, what would be necessary to do to optimize this technique, among other things, which will provide readers with a broader understanding of the subject.