skip to main content

Dealing with “Deepfakes”

Posted on 17 October 2019 by Curtis Barnes and Tom Barraclough

Have you just seen Nicholas Cage’s face on somebody else’s body? Was he singing? Dancing? You can’t quite remember him in that scene from Lord of the Rings, but perhaps you just forgot…

If you find yourself having thoughts like this, chances are you’ve just come across your first “deepfake” video.

What is a deepfake? In effect, it’s a video that makes it look like something happened when it didn’t. Of course, deepfakes can also be a photo, or an audio clip, or even a live-stream. The substance is the same – a visual or audio record that has been manipulated using sophisticated computer systems, so that the person who consumes the record sees something different to the original record. Similar, but different!

For an example, imagine this: Somebody wants to create a deepfake video of Prime Minister Jacinda Ardern playing for the All Blacks. They take a video clip from an All Blacks game, such as might be found on streaming websites like YouTube. Then they collect photographs and videos of Jacinda Ardern from the internet – this is the data upon which they will “train” the computer system. The programmer asks the computer to try and, essentially, “fix” the face of the All Black in the original video so it matches with the Prime Minister’s face. Over millions and millions of attempts, the computer gets better and better. The result is a deepfake video of better or worse quality depending on how successful the computer system was.

So what? Well, deepfakes might be comical, entertaining, or commercially valuable. But they can also be used for harmful, deceptive, or criminal purposes. According to researchers, most deepfakes on the internet today are pornographic, and often are made and published without the permission of the women who they depict. These women might be celebrities, but equally, they might be everyday people whose online photographs, videos, or audio clips have been used to produce the deepfake. These people may not know where to turn for help.

Lots of people and governments across the world are also concerned about the effects that new technologies like deepfakes might have on law and society: they can influence politics, elections, make false accusations, or even just make people feel like they have little control over their public image. It’s important to remember that deepfakes are made using technologies we already encounter in our everyday lives: digital cameras, software on our phones, social media, blockbuster films, and so on. Finding a way to protect New Zealanders’ interests without infringing their right to express opinions and exchange information is a delicate balance.

To address these concerns, researchers like us at the Brainbox Institute are working with government organisations to better prepare New Zealand for the arrival of more deepfakes and other synthetic media. We had essential support from the New Zealand Law Foundation in producing our initial report on preparing New Zealand law for deepfakes, and we currently have an Online Safety Grant from Netsafe to help us continue our work. The next step is formulating an action plan for responding to harmful deepfakes, and a roadmap for victims seeking assistance. We welcome any interest from readers.

The author with three 'fake' people. Find out more about this here.

Curtis Barnes and Tom Barraclough
Directors and Researchers
The Brainbox Institute (Ltd)

An image that shows a deepfake of Putin juxtaposed against the original video
An image that shows the Deepnude app's Twitter account on a phone

Comments

To comment as a guest without having to login, click inside the comment field below. This will reveal a field labelled "Name". Click in the "Name" field to reveal the "I'd rather post as a guest" checkbox. Tick this checkbox and then fill in your name and email address. You can use a nickname if you don't want to use your real name. Happy commenting!