Porn leads the way in deepfake innovation

Porn sites posting of “deepfake” videos have spurred concern over the use of technology to create extremely authentic looking but absolutely fake images. And nobody knows how to really stop their distribution.

This porn genre features splicing the face of a well-known person together with the body of someone else engaged in a pornographic film. While cut-and-paste fusions of still-frame pictures have a long history,  recent innovations in film-editing techniques have made it increasingly difficult to distinguish the real from the fake.

The most notable victim of deepfake pornography is the actress Scarlett Johansson. However, she is certainly not alone as there are entire websites dedicated to this type of video.

It’s not just the porn

The problem with deepfakes goes far beyond just porn and the attempted destruction of Scarlett Johansson’s life. The bigger concern is what happens when an altered video is used to provoke a violent incident or create political upheaval.

Imagine what could happen if there was a video showing armed Poles had attacked the Sender Gleiwitz radio station in Germany. Instead of posed bodies, there were be near live-action videos of the event which clearly showed marauding Poles in action. If you were the type that had to see it to believe it – that might just be enough to convince you to start a war. And yes, this was one of the instances used to spark WWII.

With advances in technology advances and easier accessibility, security experts such as Hany Farid, a computer science professor at Dartmouth College, see a “perfect storm” of deepfakes on the horizon. The only question is how soon deepfakes will be weaponized for political purposes.

The latest deepfake sensation: Mr. Bean’s on Donald Trump’s face

Here comes the law

American legislators are thinking of legislating a solution. Last month, Ben Sasse submitted a bill to the US Senate which would have criminalized the malicious creation and distribution of deepfakes. In the state of New York, a proposed state law would make it illegal to make a deepfake video without the key person’s consent.

Let’s nail the deepfake distributors

Measures against deepfakes has focused in creation and distribution. As shown by deepfake porn, the technology for making such content already accessible and in use. This shifts the focus to distribution platforms such as YouTube or Twitter.

It looks like platforms – at least in Europe – might not only have new filtering requirements stemming from copyright laws, they might also be called to police their content for deepfakes. This expanded requirement to filter content – removing copyrighted material, parodies, and yes, deepfake porn – could have a huge long-term on the shape of the internet of tomorrow.

But then, this filtering requirement also might just not be doable. Every minute, an estimated 400 hours of video are uploaded to YouTube and 350,000 tweets sent on Twitter. That’s a lot of content for platforms to be playing Whack-a-Fake with. As Scarlett Johansson put it in a Washington Post interview:  “The fact is that trying to protect yourself from the internet and its depravity is basically a lost cause.”

This post is also available in: German

As a PR Consultant and journalist, Frink has covered IT security issues for a number of security software firms, as well as provided reviews and insight on the beer and automotive industries (but usually not at the same time). Otherwise, he’s known for making a great bowl of popcorn and extraordinary messes in a kitchen.