Deepfake crimes: How Real And Dangerous They Are In 2021?
Some of you might remember the infamous Tom Cruise TikTok videos that came out earlier this year and took the internet by storm. Or the video of Obama insulting Trump that came out a few years ago. These videos are, obviously, fake and created using a technology called Deepfake, which experts think might be the most dangerous crime of the future!
You might have many questions about what a Deepfake is, how it works, and the potential danger of the technology. Well, in this article, I will answer all your questions and explain to you what a Deepfake is.
Table of Contents
A Deepfake is a form of media where any person in a video is replaced by another person that shares a form of likeness with the first. The name itself is a combination of 2 terms – deep learning (the technology used) and fake.
In simple terms, a deepfake is a technology where you can replace your face in a video with someone else's face if both of you have a few identical facial features. Theoretically, you could make yourself look like anyone else of social importance and impersonate their identity to fool the viewers of the video.
Image Source – towardsdatascience.com
While there are quite a few moral aspects and applications of the technology, deepfakes are mainly used illegally to harm important individuals such as celebrities and politicians.
According to Sensity, one of the world's first visual threat intelligence company, over 96% of the deepfakes on the internet are of pornographic content. The top 3 websites that this information is gathered from having more than 134 million views, with the most common victims being from UK, US, Canada, South Korea, and India.
Image Source – sensity.ai
But this is simply scratching the surface of the ominous implications of deepfake technology.
A well-known and impactful deepfake in the past was done by a hacktivist known as Bill Posters, who created a deepfake of Mark Zuckerberg, saying
“Imagine this for a second: One man, with total control of billions of people's stolen data, all their secrets, their lives, their futures, I owe it all to Spectre. Spectre showed me that whoever controls the data controls the future"
Even though people quickly realized that the video was fake, it made a statement of the future of deepfakes and the global implications of a well-tailored deepfake video.
Back in the early days of manipulating videos and images, it was pretty difficult to replace one person's face with another, and even once it was done, it never looked very convincing.
But with the development of machine learning and neural networks, this process has become relatively easier. An artificial neural network called autoencoder, is quite popularly used for the sole purpose of creating deepfakes.
Image Source – aajtak.in
At its core, a deepfake involves training an artificial intelligence system with multiple images of a source target, which it can then superimpose on an image of the impersonator. With more pictures, in different poses, angles, and facial expressions, that are fed to the system, the better will be the quality of the deepfake.
There are 4 main types of deepfakes – face replacement, face re-enactment, face generation, and speech synthesis.
- Face Replacement – The process of simply ‘sticking’ one person’s (source) face on another person (impersonator).
- Face Re-enactment – Involves recreating and manipulating the facial features of the source on the impersonator. Used to make it seem like the source said something that they did not in real life.
- Face Generation – This involves an artificial intelligence system creating entirely new faces with features using a pre-existing dataset of face images.
- Speech Synthesis – Creating a system capable of dictating any text in the tone and voice of the source target. This is relatively a new field in the world of deepfakes.
To be honest, creating a deepfake is as easy as downloading any free tool available online and feeding it a few hundred pictures of your source target. Once you get the impersonator to record a video saying or doing whatever you want the source to seem to do, you are all set to create the deepfake.
With the possibilities of the deepfake techniques and the ease of access to these, their area wide variety of applications for these techniques as well. The applications vary from simple face swaps for pornographic content to complex speech synthesis and face replacement of political leaders.
Image Source – the-sun.com
Much of the deepfakes on the internet involve pornographic content of people, quite often females or female celebrities, generally without their consent. This usually involves creating the source person's head over that of an actor in a video of pornographic nature.
Some of the earliest deepfake pornographic content hit the surface web in 2017 through the popular website, Reddit. One of the first such celebrity deepfakes that caught the general public's attention was that of Star Wars actress Daisy Ridley, which initially created a lot of buzz in the industry.
Over the years, more and more women have become the target of such non-consensual fake pornographic videos, with such websites gaining almost 130 million views in just the past year.
In her paper “Beyond ‘Revenge Porn’: The Continuum of Image-Based Sexual Abuse", Professor Clear McGlynn explores various possibilities of using such images for revenge and makes an argument for classifying this as sexual abuse, among others.
Image Source – bbc.com
The possibilities of using deepfakes in politics are quite limitless as well. While most of it isn't officially created or released by any political parties, they do cause quite a lot of buzz in the political landscape.
In April 2021, a couple of Russian men managed to use deepfake technology to impersonate Russian opposition leader Alexei Navalny. They managed to get into various sensitive conversations and even interviews on popular Latvian media.
The impersonators even managed to arrange video calls with parliaments from UK, Latvia, Estonia, and other European countries. The pranksters later went on to laugh of the incident claiming it to be a prank. They also denied various allegations linking them to the Kremlin, but it is likely that their work is useful to and supported by the Russian government.
Similar issues have happened in the US as well in the past, involving ex-president Obama and ex-president Trump where the former seemed to verbally insult the latter.
Such incidents that might seem funny in hindsight, however, have the potential to turn the world upside down by causing threats to security and such.
Image Source – disneyplus.com
With the possibilities of manipulating and tailoring videos using deepfakes, it comes as no surprise that the technology found its way into the movie and art industries as well.
Deepfakes in movies have been used mainly for humanoid characters, de-aging, and bringing dead actors back to life. The latest venture in the deepfakes in the industry was Mark Hamill’s came as young -Luke Skywalker in the season 2 finale of the Disney tv show, The Mandalorian.
You and I know how good it looked when Luke came in wielding his lightsaber after fighting the Dark Troopers. A YouTube channel called Corridor Crew in their videos explores just this and explains how the entire system works.
Art on the other hand, uses deepfake for a different purpose; Many artists try to bring their art to life digitally using deepfakes. In March this year, a genealogy platform called My Heritage released a tool named Deep Nostalgia. You can scan old family photographs and bring your old relatives back to life.
The Deep Nostalgia tool is straightforward to use, involving simply visiting their website and uploading the image of a person you want to bring "back to life". It would be nice to show your mum and dad pictures of their parents who might have passed away early.
Image Source – i-sight.com
Identity theft has been an issue globally ever since the start of the information age in the late 20th century. While the rewards for identity theft have grown in this day and age, so has the technology.
Deepfakes make it possible for malicious actors to even impersonate the identity of dead people. They can start new bank accounts, access pre-existing accounts, and even apply for new credit cards using this stolen identity by utilizing the deepfake.
Just last year, over $3.4 billion was reported as losses due to such fraudulent activities using the deepfake technology.
Even though most banks are switching to voice recognition systems, hackers are able to override it due to the recent developments in the speech synthesis aspect of the deepfakes.
Recently in spring 2021, a couple of fraudsters in China (last names Wu and Zhou) were arrested for committing tax frauds by assuming false identities using the deepfake technology. According to Chinese police, the duo had been issuing fake tax invoices to customers since 2018 and had managed to make over $76.2 million with the fraud.
While nations don't seem to provide much for the protection and prevention of deepfakes, various social media platforms seem to have taken matters into their own hands against deepfaked media of harassment or misinformation nature.
Image Source – slate.com
Twitter recently started active countermeasures against such deepfaked content on their platform. To keep everything in control, Twitter started placing a little notification along with such manipulated fake media, informing the user of its non-authenticity and it is being fake.
Twitter also holds the right to remove any tweets that contain such manipulated or deepfake media if it violates any regulations.
In contrast to that, Facebook has actually begun encouraging the creation of such deepfake media. They expect to develop the technology so far so that they can create a state-of-the-art deepfake detection software. They prominently hosted the Deepfake Detection Challenge in 2019, with over 2000 participants generating over 35000 detection models.
Image Source – facebook.com
Following in the footsteps of Twitter, the platform also decided to take down any content created using artificial intelligence to manipulate the media to spread misinformation or harassment.
Deepfakes, for all the negative press they create, aren’t entirely without any positives. They can be used to give a better experience in the field of movies and art, and also for entertainment and memes. While not all applications might be practical as of now, they definitely do make the trade-off a bit better.
In this article, we explored the concept of deepfakes, their working principles, applications and finally wound up discussing some platforms fighting against the negative impacts of deepfakes.
My verdict on deepfakes is that as many negative impacts it has caused, if appropriately regulated, it can be very beneficial in the entertainment industry as well as various other fields. It isn't to be cast out labeling it a potentially dangerous crime.