top of page
Writer's pictureKhalid Amin

Types of deepfakes and how they are made.

Updated: Aug 26, 2021


Types of Deepfakes

There are two main types of deepfakes, deepfaces, and deepvoices. Let's analyze each of them.


Deepface

Deepface involves creating convincing but completely fictitious photos from scratch.

Image animation aims to generate video sequences so that the person in the source image is animated according to the movement of a video. The goal is to create fake videos.


This technology is within the field of computer vision, and AI researchers have been working to produce more realistic videos. It leverages machine learning to manipulate and generate images or videos that replace one person with another.


Deepvoice

Deepvoice is the impersonation of a person's voice in audio, making it sound like their real voice.


Until recently, if we told you this, you would think we were crazy. But with AI-driven synthetic media like deepfakes on the rise, this is already starting to happen.

Take, for example, the world's first AI-driven cybercrime reported in early September 2019. Using speech synthesis technology, thieves were able to convince an energy executive to think he was on the phone with the CEO of his parent company, tricking him into transferring more than $250,000 into his account.


How is a Deepfake constructed?

University researchers and special effects studios have pushed the boundaries of what is possible with video and image manipulation.


But deepfakes were born in 2017 when a Reddit user of the same name posted fake porn videos on the site. The videos swapped the faces of celebrities (Gal Gadot, Taylor Swift, Scarlett Johansson, and others) for porn performers.


Making a face-swapping video is quite simple. In addition, there are deepfakes programs.


First, run thousands of facial shots of the two people through an AI algorithm called an encoder. The encoder finds and learns similarities between the two faces and reduces them to their shared common features, compressing the images in the process.


A second AI algorithm called a decoder is then taught to recover the faces from the compressed images. Because the faces are different, it trains one decoder to retrieve the first person's face and another decoder to retrieve the second person's face.


To perform the face-swapping, simply input encoded images into the "wrong" decoder. For example, a compressed image of the face of person A is fed to the decoder trained on person B. The decoder then reconstructs the face of person B with the expressions and orientation of face A for a convincing video.


Another way to do deepfakes uses what is called a generative confrontation network, or Gan.


A Gan pits two artificial intelligence algorithms against each other. The first algorithm, known as a generator, receives random noise and converts it into an image. This synthetic image is added to a sequence of real images, of celebrities, say, that are fed into the second algorithm, known as the discriminator.


At first, the synthetic images will not look like faces. But if you repeat the process countless times, with feedback on performance, the discriminator and the generator improve. Given enough cycles and feedback, the generator will begin to produce completely realistic faces of completely non-existent celebrities.


Why are they a threat?

The most insidious impact of deepfakes, along with other synthetic media and fake news, is to create a zero-trust society, where people can't, or no longer bother to, distinguish truth from falsehood. And when trust is eroded, it is easier to generate doubt about specific events.


As technology becomes more accessible, deepfakes could spell trouble for courts, particularly in child custody battles and employment tribunals, where false events could be entered as evidence.


But they also pose a personal security risk: deepfakes can mimic biometric data and can potentially fool systems that rely on face, voice, or vein recognition. The potential for scams is clear.


Call someone you don't know and they are unlikely to transfer money to an unknown bank account. But what if your "mother" or "sister" sets up a video call on WhatsApp and makes the same request?


Deepfakes are a major threat to our society, political system, and business because:

  • They put pressure on journalists who struggle to filter real news from fake news.

  • Threaten national security by spreading propaganda and interfering in elections.

  • Hinder citizens' trust in information from authorities, and raise cybersecurity issues.

  • Raise cybersecurity concerns for individuals and organizations.

26 views

Comentários


bottom of page