Patrick Johnson
As we stated back in a Veriff blog in August, deepfakes are one of the biggest fraud trends out there at the moment. Worryingly, this questionable practice isn’t just being used for facial recognition fraud but is seeping into many other facets of life as well.
So, what really is a deepfake?
The term refers to a falsified digital representation of someone that has been created by artificial intelligence. It could be a video, photo or audio recording which has been manipulated to such an extent that it appears real. Deepfakes are usually used to mislead people in a worrying manner although there are plenty of spoof, satire and simply mischievous creators out there whose intentions aren’t as dark as you may think.
As a perfect example of a tongue-in-cheek representation that shows you precisely how good deepfakes can be, take a look at the below video made by Trey Parker and Matt Stone, creators of South Park.
In January, biometric authentication technology supplier iProov released some data following a survey of more than 100 security decision makers in the financial services sector and their thoughts on the threat of deepfakes.
They found that over three-quarters (77%) of cyber security decision makers are worried about the potential for deepfake technology to be used fraudulently and 71% of respondents said they thought their customers were at least somewhat concerned about the threat.
And a perfect demonstration of the incredible growth of deepfakes online is shown in a recent report by Sentinel, who mention "Since 2019, the number of deepfakes online has grown from 14,678 to 145,227, a staggering growth of ~900% YOY."
In a move that reinforcing that deepfakes have become more commonplace, Facebook announced new policies in January that ban deepfakes from the platform.
Here at Veriff we’re keeping a close eye on this trend and are ensuring that our practices around facial recognition and identity verification security are as robust as possible.
How are deepfakes created?
As you might expect, special effects artists and academics have long explored how to achieve new heights in video and image manipulation. This certainly provides the backdrop for deepfakes but when was the term first coined?
In 2017 a Reddit user posted a series of fake pornography clips onto the site showing the faces of female celebrities superimposed onto porn performers. This is now used as the example as when deepfakes as we know them today were first created.
In its simplest terms, in order to produce a deepfake, you need to be able to ‘face swap’ two people using a decoder. The final video is therefore comprised of the face of one person, with the expressions and orientation of another. To make it a realistic, every frame has to be transposed in this way.
These days, programs called generative adversarial networks (GANs) are used to combine two artificial intelligence algorithms. When the process is repeated several times, the synthetic image that is created becomes a wholly believable representation of either a well-known person or totally fabricated persona.
Who makes deepfakes?
It appears as though the trend for deepfakes has become extremely prevalent, with professionals from many different industries now dabbling in the technology. Academics, amateurs, hackers, visual effects studios, TV producers, comedians and even governments are creating and using deepfakes.
In short, anyone these days can download deepfake software from the comfort of their own home and produce pretty convincing videos or images as a side-hobby.
How do you spot a deepfake?
As with most technology, as the practices have become more complex and refined, deepfakes are becoming increasingly harder to spot. As our methods to identify them improve, so does the technology being used to create them – hence a seemingly endless battle.
For example, in 2018, US researchers announced that deepfake faces could be detected because they didn’t blink properly, but no sooner had this been announced, deepfake videos began to emerge with a new ability for blinking.
Some of the things to look out for are as follows:
- Poor lip syncing
- Patchy skin tone
- Flickering around the edge of the face
- Hair strands out of place
- Badly rendered jewellery
- Badly rendered teeth
- Inconsistent lighting effects
- Incompatible reflections on the iris of the eyes
Just like the team at Veriff, governments, organisations, and tech firms around the world are currently working round the clock to research ways to detect deepfakes and prevent them from being used fraudulently.
Why are deepfakes dangerous?
It doesn’t take a genius or a tech guru to understand why deepfakes could be dangerous or used for nefarious purposes. The concept of anything being faked can be unsettling but when you are talking about someone’s identity it becomes altogether more sinister.
While some deepfakes are created just to garner amusement or to emphasise a point, it can be surmised that the majority of them are intended to create harm.
Disturbingly, one of the biggest impacts of deepfakes is that they can be used to create fake news and thus a zero-trust society where truth cannot be differentiated from falsehoods. This creates an environment where people don’t know what to believe or who to turn to for the truth.
Professor Lilian Edwards, a leading expert in internet law at Newcastle University in the UK described this as such to UK national paper The Guardian: “The problem may not be so much the faked reality as the fact that real reality becomes plausibly deniable.”
This implications of this are far-reaching, with many different industries and scenarios potentially under threat:
- Court cases - Deepfakes could be used to falsify evidence for things like custody battles and employment tribunals
- Security - If deepfakes are used to mimic certain biometric data then it has the potential to trick security procedures on secure accounts, making robust systems like Veriff’s absolutely essential
- Scams - A deepfake audio file could be used to trick someone into revealing personal information or transferring large sums of money
- Political attacks - The trustworthiness of politicians or organisations could be called into question if they are the victim of a sophisticated deepfake
In recent years, high profile figures and celebrities have become regular victims of deepfakes due to there being so much footage of them available in the public space. However, as the number of photos and selfies the average person takes increases, it is only a matter of time before deepfakes are soon making their way out into the wider world.
Examples of deepfakes
- Movies: Hollywood has long been transposing real or fictional faces onto other actors. In 2016, Rogue One: A Star Wars Story brought Peter Cushing back to life for several scenes.
- Pornography: In September 2019, the AI firm Deeptrace announced that they found more than 15,000 deepfake videos online, with an incredible 96% of these being pornographic.
- Politics: American actor, director and producer Jordan Peele and Buzzfeed created a video of Barack Obama apparently calling President Donald Trump “a total and complete dipshit.”
- Fraud: One of the best-known examples of an audio deepfake occurred in 2019 when the chief executive of a UK energy firm was tricked into sending €220,000 ($240,000) to a Hungarian supplier. He was apparently instructed to do so by the CEO of his company’s parent firm in Germany, but it turned out to be a deepfake and the perpetrators were never caught.
Do you now feel clued up on deepfakes, what they are and their current applications? Take this quiz to test your knowledge!
If you want to fight deepfakes, and stop account takeovers, you can use Face Match, Veriff's dedicated reverification product to make sure it's really your customer signing in from a new country.
