investigation
How ‘deep fakes’ pose a risk to your safety, elections and the share market, according to cybersafety experts

Tom Cruise stares deep into the camera, gives one his trademark chuckles and tells the social media audience on TikTok: "Just wait 'til you see what's coming next". Everything about it is classic Cruise. The good looks. The cheeky smirk that has grossed more than $8 billion at the Hollywood box office. Even the sunglasses and the confident swagger as he gets up and drives a golf ball off a tee on a perfectly manicured patch of grass. There's only one catch. This isn't the guy known for playing for Ethan Hunt in the Mission Impossible franchise. This isn't the actor known for Jerry Maguire or Risky Business. This isn't the Scientologist who once leapt off Oprah's couch. It's a fake. Technically, it's what is known as a "deep fake". And it's going to become a lot more common. Deep fake is much more than changing your lips on a face swapping app. It’s sophisticated. It looks real. It’s the ultimate hack. Deep fake is a doctored video or image using Artificial Intelligence, which can make a person appear to look like someone else, as well as do or say things they never did. And many warn it is poses a grave 21st Century threat the world. FICTION NOW FACT Three decades ago, before the term had been coined, deep fakes could only be found in the movies. And we weren’t being fooled, we were being entertained. You may remember the 1987 film, The Running Man, where Arnold Schwarzenegger’s character is framed for a crime he didn’t commit after edited footage makes it look like he opened fire on a crowd of protesters. Later, his oppressors manipulate existing vision to digitally insert his face onto another person’s body to make it look like he was killed. Another example is the famous 1985 movie Weird Science where two teenagers design their ideal woman on a computer with a freak electrical accident bringing her to life. But put the internet, digitisation, AI, machine learning, sophisticated online editing and social media together and the future has arrived. Now there’s fears synthetic images, video, audio and text are or will be used by state-sponsored hackers in Russia, China and other countries in disinformation campaigns that are increasingly difficult to detect and expose. CyberCX Chief Strategy Officer Alastair MacGibbon said deep fakes risked accelerating malign trends in a conspiracy-minded society where people could no longer be sure what they saw with their eyes or heard with their ears was real anymore. “Imagine a world where we just can’t trust, which is what criminals want,” Mr MacGibbon said. “It shows you the potential of badness. Unfortunately, every time you think something is a fiction, someone proves to you that they’re going to do it.” Mr MacGibbon said in the past 18 months deep fakes had become so believable and dangerous they now posed a fundamental risk to politics or free markets. “I suspect in time you will see people who want to disrupt elections using deep fake technologies to change voter behaviour very close to the time they vote,” he said. “Someone could even put out a video pretending to be Elon Musk talking up Bitcoin just after they’ve bought it, and then they sell the Bitcoin when it’s at the top of the market, showing how easy they can manipulate markets. “People are also faking phone calls where there’s enough of your voice already out there and then they call your family using your voice to ask for money because you’re locked out of the house with your wallet inside and can they send it to this bank account.” Last week, a new documentary on the life of Anthony Bourdain came under fire for doing just that — using AI to create the voice of the late American food critic long after his death and then manipulating it to have him “narrate” in the film. The film was originally praised for its authenticity before director Morgan Neville revealed how he had done it, telling GQ Magazine about 10 hours of Bourdain’s recorded speech had been fed into a machine to create the most realistic narrative voice possible. “The bigger the quantity, the better the result,” he said. “We also had to figure out the best tone of Tony’s voice – his speaking voice versus his narrator voice, which itself changed dramatically over the years.” Fans expressed feelings of anger and betrayal when the deception was revealed. PHOTOSHOP ON STEROIDS The term deep fake was coined in 2017 by a user on American discussion website Reddit. He would share deep fakes he created with the Reddit community, many with the faces of celebrities swapped onto the bodies of actresses in porn videos. This insidious new digital trend took off, with users generally delighting in the combination of malice and creativity. Pre-existing footage of Kit Harrington’s character Jon Snow in Games of Thrones was doctored to make him apologise for the show’s final season. Donald Trump was manipulated to read a book of Rudolph the Red-nosed Reindeer. A video of Mark Zuckerberg was posted on Facebook-owned Instagram showed him boasting about how he “owned” users. It was a protest at Facebook’s refusal to take down another doctored video of top Democrat Nancy Pelosi. A crude attempt was of a 2020 election video of what appears to be Queensland Premier Annastacia Palaszczuk giving a press conference, criticising her own Government and saying that she would understand if voters “want to get rid of us”. In another North Korean dictator Kim Jong-un was featured extolling the virtues of demoracy — this one admitted it was fake and was the work of a political reform lobby group. But it’s not just celebrities and politicians who are targeted, false images of everyday Australians are also doing the rounds online after getting into the hands of criminals. WA’s Noelle Martin found herself in the same situation as English actor Daisy Ridley whose image has been superimposed in pornographic content, called image-based abuse or revenge porn. She had her social media photos lifted and doctored to become pornography in 2018. Some of the images of Ms Martin were of her as a 17-year-old and she says the impact has been devastating. “I’m just an ordinary person. I’m not a celebrity. I’m not a public figure,” Ms Martin said. “It was extremely violating and degrading. They were taunting me.” The 26-year-old does not know who the perpetrators were and is trying to stop other innocent women from going through the same harm and humiliation. “It’s extremely gendered. It’s predominately women,” she said. “It’s upended my whole life in every way. No law firm wanted to hire me. It’s extremely difficult to find employment. The impacts on romantic relationships. Your future children and family will know about this. It’s so permanent and ruins your whole image in such a violating way. “I want the ability to be able to use social media, wear what I want to wear, express my identity, and take part in society without being abused and that’s the standard that I’m fighting for, not that we have to control every we put out there because this is what might happen to you.” About 93 per cent of deep fakes are currently pornographic. Some victims are blackmailed. But often the aim is simply to embarrass, hurt, humiliate or bully. And it is hard to prove it’s not real. As technology becomes more advanced, the creation of deep fakes has become commercialised with the development of apps such as FakeApp and Faceswap. “We’re going to keep seeing these types of apps because people enjoy doing this sort of stuff,” Edith Cowan University Computing and Security Associate Professor Paul Haskell-Dowland said. “Not only has the software got better but it’s become easier to use and is available to everyone. There are tools you can download for free. The barrier to entry is almost zero now.” CLASSROOM FEARS Academics and cyber safety experts are worried about what deep fakes mean for the next generation of cyberbullies, not dismissing the fact it could already be in use by children in the schoolyard. “If any group is going to be vulnerable to using this, it’s today’s teenagers because they have grown up with technology, they’re immersed in social media, and the tools are only going to get easier,” Associate Professor Paul Haskell-Dowland said. “The intent is quite innocent, although the consequences can be much more severe.” Cyber safety expert Susan McLean said education was the key to lowering the risk of harm among students. “We know that a lot of content that young people are exposed to is not suitable for them,” Ms McLean said. “There’s the possibility that they will try things out because they’ve seen it online. “What I have seen is someone taking a photo of someone and then putting their head of someone else’s naked photos so you could probably call that a deep fake. That is something I’ve been dealing with for many years in schools.” Ms McLean said cyberbullying was the number one non-academic issue Australian schools dealt with. The eSafety Commissioner — a Federal Government platform to help Australians experiencing online bullying or abuse by taking action — is yet to receive a report of cyberbullying involving deep fakes but says it is only a matter of time. “As deep fake technology is becoming increasingly accessible and sophisticated, we anticipate that everyday Australian citizens will become victimised by this threat vector,” Commissioner Julie Inman Grant said. “In fact, our image-based abuse complaints scheme will cover reports of deep fakes and we’re girding ourselves for reports.” COMBATING DEEP FAKES Ms Grant said combating deep fakes had “become a technology arms race” with companies having a responsibility to engineer out the risks of misuse. “Innovations to help identify, detect and confirm whether content is authentic are also advancing,” she said. “A number of industry and government initiatives are under way, with most focusing on flagging content if there is a high probability that it has been tampered with. “Blockchain technology, digital watermarking, digital signatures and other tracking and tracing tools are all being deployed and tested as possible deep fake detection solutions.” While we fear deep fakes are everywhere another concern is whether it will become the new "my twitter got hacked" excuse for wrong-doing. False cries of “deep fake” to muddy the waters threaten to be just as damaging to public discourse in politics and to sow even more division. Donald Trump’s ‘fake news’’ brush-off was a precursor. “What will happen is, a politician will do the wrong thing, make a racist or sexist comment, and they’re going to say it’s a deep fake,” Mr MacGibbon said. “People are going to undermine accountability and say it was fake and for others to prove it was real.” For more information or to report a deep fake: esafety.gov.au.

investigation
How ‘deep fakes’ pose a risk to your safety, elections and the share market, according to cybersafety experts

Subscribe today.

Digital

Everyday Digital

Digital & Print

Weekend Papers + Everyday Digital

investigationHow ‘deep fakes’ pose a risk to your safety, elections and the share market, according to cybersafety experts

Subscribe today.

Digital

Everyday Digital

Digital & Print

Weekend Papers + Everyday Digital

investigation
How ‘deep fakes’ pose a risk to your safety, elections and the share market, according to cybersafety experts