Deep Fakes – Don’t Trust Everything You See

You are currently viewing Deep Fakes – Don’t Trust Everything You See

Have you seen the video of David Beckham campaigning for malaria awareness in nine different languages, Obama calling Trump a complete dipshit? If the answer is yes, then you have seen a Deep Fake. Sometimes it is impossible to distinguish between deep fakes and real videos.

In this article of Terminal Stack, we will learn about Deep Fake contents and how we can identify them. 

What are Deep Fakes? 

Deep Fakes are the fake content in the form of videos, pictures, or audio created with the help of machine learning and face-swapping. You can call them 21st century’s Photoshop.

This term ‘Deep Fake’ came from deep learning algorithms that train themselves how to solve problems when given large sets of data. This helps to swap faces in the video and create a new video that portrays statements or actions that never actually happened.

Deep fake media also includes deep fake audio which uses the same technology.

Also Read: Artificial Intelligence and its scope

How do the Deep Fakes Work?

We are familiar with the functions of apps like Snapchat which use face swaps or filters to change our facial expressions. Deep fakes are just like that but with more use of technology.

These deep fakes are created with the help of a machine learning technique called Generative Adversarial Network (GAN). It uses two AI algorithms, the first creates the fake content and the second detects and improves any flaws.

GAN can also create computer-generated images of fake humans. These fake human images have been used by a website called ‘This Person Does Not Exist.

There are several apps that make it easy even for beginners to create deep fakes, such as the Reface app, FaceApp, or Face Swap app.

This is a Deep Fake Video

What threats Deep Fakes pose?

Deep fake technology is relatively new, but it is improving very quickly. It has become very difficult to distinguish deep fakes from real videos.

There is a fair explanation by US Senator Marco Rubio about how deep fakes can pose a threat to democracy. He said, “In the old days, if someone wanted to threaten the United States, they needed 10 aircraft carriers and nuclear weapons, and long-range missiles but now all they need is the ability to produce a very realistic fake video that could undermine our elections, that could throw our country into tremendous crisis internally and weaken us deeply.”

As we have seen, there are many apps that are easy to use and free of charge allowing ordinary people like us to make face swaps and use them to harm the character of anyone we want, to frame them for wrongdoing, or to create revenge porn like stuff.

Out of the total deep fake videos found online, pornography made up 96% according to the Deeptrace report.

There could be many corporate scams where deep fake audio is used to ask an employee to send money pretending the person on the other line is their CEO. 

In the East African country, Gabon, a deep fake video led to an attempted military coup because of a misunderstanding.

But there is a positive side too – 

This technology can make language barriers a thing of the past as we have seen in the video of David Beckham about the anti-malaria campaign where he spoke in 9 different languages.

Sometimes the actor passes away during the shooting of a film or sometimes a character needs to be older or younger than the actor himself, here deep fake technology can fill the place. 

David Beckham speaking in 9 different languages.

How to spot Deep Fakes

We can easily identify poorly made deep fakes but higher quality deep fakes are difficult to spot.

Some characteristics to spot deep fake videos are –

  • Unnatural eye movement.
  • Unnatural hair.
  • Unnatural facial expressions.
  • Unnatural body shape.
  • A lack of eye blinking.
  • A simple stitch of one image over another.
  • Different skin colors.
  • Robots like voices.
  • Awkward head and body positioning.
  • Contrary head positions.
  • Different lighting.
  • Blurry or misaligned visuals.
  • Bad lip sync.
  • Digital background noise.

Facebook and Twitter have banned deepfake videos. Their policies involve taking down a video as soon as they detect it as a deep fake.

In 2019, Facebook hosted the Deepfake Detection Challenge for the creation of new technologies to detect deep fakes. The competition had prize money of $500,000.

A program like Deeptrace is a combination of antivirus and spam filters that detects the media and quarantines mistrustful content. 


Here are some ideas to spot deepfakes on a personal level –

If we’re watching a video online, it’s a good practice to confirm that the video is from a trusted source before believing its claims.

When you receive a call from someone like your boss, always make sure the person on the other end is really your boss before taking any action.

Don’t believe everything you see on the web. If you think it’s unbelievable, it may be. 😉

References :

Leave a Reply