Can you imagine that someone can impersonate your identity with Artificial intelligence? Recently there has been talk of the phenomenon of deep fakessomething that has worried Famous and familiar faces. Nowadays, it seems more and more difficult to detect if the content spread by the network is true.
If you also have doubts about whether what you see on the Internet is real or not, keep reading this post where we tell you What are the deep fakes and how to detect them.
It may interest you: Postgraduate in Artificial Intelligence and Machine Learning
Application of Artificial Intelligence in Fintech
What are the deep fakes?
Recently, a video of actress Jennifer Lawrence with the character of actor Steve Buscemi appeared on the networks. At the same time that Lawrence spoke in a speech in front of the cameras, the actor’s face articulated the same expressions as hers. The video is a clear example of how fake news is no longer just text. We are facing the development of synthetic images, better known as deep fakes.
The deepfake number comes from deep learningtranslated as deep learning, one of the currents of Artificial Intelligence. In this case, it is learning with artificial intelligence that to be used with the intention of creating false content.
In most cases, this is based on video manipulations in the software analyzes the source material and extra part of the, then insert and fit it into another video. The so-called face swaps or deep video portraits It is the most common form of false.
It seems more and more difficult to detect what is real and what is not. This does not mean that it is impossible, even deep fakes best worked can be detected by specialists in style transfer techniques.
In an interview with Diario El País, Artificial Intelligence specialist Jorge Muñoz assured that we are facing a process that could scale beyond the world of celebrities. “The biggest problem with these techniques is that they require a large amount of data to be able to train them, so today we will only see content like this about famous people who can find millions of hours of high-quality video on the Internet.”
Deep fakes are not only videos
The videos deep fakes they have become the most widespread technique for creating fake news (also known as FakeNews) and the data proves it. According to the I Study on the impact of Fake News in Spain prepared by the Complutense University of Madrid:
- 86% of the Spanish population creates fake news Wow fake news.
- 60% of Spaniards think they know how to detect them, but in reality it is 14% who differentiates it.
- 4% of the population create and distribute fake news
However, they are not the only means of counterfeiting. The technology fake deep allows you to create completely fictional photos from scratch. Cases such as “Katie Jones” have been known, a profile that confirmed working at the US Center for Strategic and International Studies, but which, it is believed, was created to carry out espionage operations abroad.
The same happens with audio. Deep fake allows you to create “voice clones” of public figures.
How does the fake deep?
through a Artificial intelligence algorithm called encoder thousands of shots of the faces of the people who want to be faked are executed.
What the encoder does is identify and learn the similarities between the faces and the reductions to shared common features through a process of image compression. Then came into operation un second algorithm called decoder, which promises to recover the compressed images.
To perform the face swap, the encoded images are introduced in the opposite decoder so that it reconstructs the face of the other person. That is, an encoded image of face A feeds the decoder input into face B. The decoder is responsible for reconstructing face B with the expressions and orientation of face A to obtain a convincing image.
How to detect the deep fakes?
As technology evolves it has become more difficult to detect what is true and what is not. Below you already have some features that you can try to detect a fake deep
Humans typically blink once every 2-8 seconds, with each blink lasting between 1 and 4 tenths of a second. If you want to know if you are facing a DeepFake, look at the number of times it blinks, you realize that he does much less than a real person. The Loss Algorithm DeepFake he cannot blink at the same speed as a human.
face and neck
Does the body conform to the face where the posture matches the facial expression? Most of the DeepFake their main facial substitutionschanges in the body can only be implemented with great effort.
If the body of the person who has applied a face has other features that do not match, for example tattoos, pronounced muscles or a different skin color, it is a sign of forgery.
Although the technology is already very easy to use, the process of learning to make fakes are very worked. Therefore, most of the deep fakes that are shared only a few seconds in duration. So if a very short clip of implausible content needs to be verified and there is no obvious reason why the recording is so short, it is often because it is a false.
In this context it is important because the capture leak, to the person who first compared the deep fake video. a often, andThis helps to plug in the context of the post already check if the source material was more detailed after all.
sound for recording
Not only the image exposes a DeepFake, but also the sound. The software is often limited to changing the picture, but not adjusting the sound. sentences, if the sound is not present or does not match the imagefor example, with a poorly implemented lip touch, this indicates a fake.
Ten in mind the details. By verifying the content of the video, also it is useful to let it play at half speed. For example, small discrepancies in a person’s background or sudden changes in the image that are immediately noticeable.
inside of mouth
The software to create deep fakes has been able to transfer faces pretty well so far, but the devil is in the details. For example, some blur inside the mouth is another sign that it could be a fake image. Artificial Intelligence is still struggling to correctly represent the teeth, tongue and oral cavity when speaking.
positive uses of fake deep
Apart from maintenance, there are different applications that can be useful with this technology. One of them has to do with restoring the voice of people who have suffered from illnesses that have left them mute. Let’s look at some other applications:
- art and entertainment: In Florida, the Dalí museum has a deepkafe of the painter that presents his art and takes selfies with visitors. In the same way, the producers of the film did RogueOne from star Wars in 2016, bringing actor Peter Cushin to life.
- Research: The generative technology it is already allowing results to be seen in the medical field, through the use of fake brain scans based on real patient data. This is a useful technology to detect possible tumors.
As Artificial Intelligence evolves, specializations in this sector grow more and more. Companies are looking for up-to-date professionals in the sector, so if you want to work to be part of the change, join the Postgraduate in Artificial Intelligence and Machine Learning. What are you waiting for!
Postgraduate in Artificial Intelligence and Machine Learning
Learn about Big Data, Machine Learning and Deep Learning
I want to sign up!