In the modern world, technology and creativity sometimes mix in surprising ways, leading us down unexpected paths. This is where deepfakes come in, a kind of ‘digital double’ that can be so convincing that it tricks us into thinking we are seeing or hearing someone real.
With the evolution of Artificial Intelligence and digital manipulation techniques, this virtual reality is increasingly present, becoming even worrying. But not everything is lost. We have an unexpected hero in our fight against deepfakes: biometrics. Biometric technology provides tools to detect these digital impostors and protect our online identity.
So sit back, as you are about to learn exactly what deepfakes are and how biometrics can help us stop this cyber threat.
Índice
What’s a deepfake
Deepfake is an acronym formed by the English words “fake” and “deep learning”, one of the currents of Artificial Intelligence. It’s a video, image or audio generated to mimic a person’s appearance and sound. They are artificially generated, and are so convincing, so realistic, that the human eye often doesn’t perceive that it is in front of a fictitious image.
The most common are face swaps, also called deep video portraits, in which the system analyzes the source material and extracts part of it to insert and adapt it into other material.
How deepfakes work
The systems that create deepfake are quite complex. They use Generative Adversarial Networks (GAN), which functioning is based on two principal matters.
The first point is that it’s a technology that understands pretty well how human faces are, and learns quickly how to take those attributes to another face. The second point is that they’re built on pieces that work as opposed forces. What does this mean? That while a portion of the system creates visibly phony information, another part is trained to point them out. As a result, a deepfake system is both its own coach and professor at the same time.
That’s why the first deepfakes were so bad–they were actually very easy to identify. But as the machine learning technology evolved, those results became more and more realistic and difficult to detect. But don’t worry! We can fight this threat.
The role of biometrics against deepfakes
Biometrics is a strong ally against deepfakes. Why? Let’s see.
- Each person is unique–they have unique eyes, noses, lips… Biometrics can identify those unique characteristics and compare other faces with the original face. It uses patterns to do so, and prevent fake faces from logging in or signing up. And of course, a biometric system can identify the contour lines around the face, which usually don’t match when looking closely.
- Biometrics, however, analyzes other elements beyond the distance between the eyes or the color of the eyes. It also considers lighting and depth. And, if it’s a video, it takes into account the subtle movements we make, such as blinking or breathing. In deepfakes, it’s common to see hair and lip movements out of line, as well as unnatural blinking or a lack of emotion that doesn’t fit the conveyed message.
- One of the hardest things to do is to verify someone’s identity by recognizing their voice. Why? Because our voices change throughout the day, unlike our face–we might sound one way when we’ve just woken up and another way when we’re at a party. But synthetic voices have small, little giveaways that can be found by voice recognition systems.
Deepfake detection techniques
Did you know that 96% of deepfakes are detected thanks to biometrics? This proves the enormous power your characteristics have to protect your digital identity.
Biometric technology uses features such as your face structure, voice patterns or your iris, which are unique for each one of us, to authenticate someone’s identity. Through sophisticated algorithms and automatic learning techniques, the system analyzes the features and looks for inconsistencies that are typical of an imitation.
Subtle but clear differences on the blinking rhythm or on the lips’ movement can be the key to detect a manipulated video.
See all posts 
