The history of deep fake technology is surprisingly long. Researchers at academic institutions have been developing deep fake tech since the early 1990s. The idea is even older, as popular science fiction—like the 1987 film The Running Man—can attest. But deep fakes are no longer relegated to the realm of sci-fi; they are, in fact, more present in our daily lives than you might realize.
As much as we would like to think that art, acting and memes are the main purposes of deep fakes, it’s difficult to overlook things like blackmail and sockpuppets. These bring into question the potential impact of deep fakes in politics and across social media.
Deep fakes are so effective that they challenge the notion that “seeing (or hearing, or reading) is believing.” At Black Hat USA 2021, Matthew Canham (CEO of cybersecurity consultancy Beyond Layer 7) presented a talk on how deep fakes could be used to facilitate social engineering attacks and fraud. Some of the examples used are scams that we’ve already seen play out in the real world. Matthew hilariously used the example of the “I’m not a cat” lawyer to explain the possible future of deep fake scams and to demonstrate the reality: We already have the technology and this future is inevitable. So, what, exactly can we expect to see going forward, and what can we do about it?
We’ve already witnessed deep fake former U.S. president Barak Obama make a public service announcement, deep fake Mark Zuckerberg discuss online privacy and several actors appear in … compromising pictures and videos that never actually happened. These fooled a lot of people.
As deep fake technology improves, it will become even faster and more advanced. The ultimate deep fake will be a live video backed by an AI bot that can interact with participants (read: Victims) in real-time. Similar bots can already interact in chat applications and in email threads; responding to statements and inquiries in a manner consistent with the individual they are impersonating. Combine that with, say, a real-time video in a Zoom chat, and the worst outcome won’t just be a coworker accidentally leaving the cat filter on. Instead, there could be a deep fake version of your CEO or another trusted person asking for account credentials, money or other potentially damaging information.
Let’s revisit the video of president Obama for a moment. This turned out to be a rendering of the former president based on a video of comedian and actor Jordan Peele. FakeApp was used to process and refine the footage to make it believable. High-profile politicians and other public figures are prime targets for this type of attack, as there is often ample footage available to train the backend AI on.
With improving technology, a realistic use for deepfakes today is similar to the cat-lawyer. It’s already possible to generate renderings of particularly vocal and media-friendly CEOs—someone like Elon Musk or Dan Price—and a “live” video could feature an actor playing the visual role, while their voice is replaced with the one you’re familiar with.
Deep fake technology is here to stay, and it’s only going to get harder to discern from actual, legitimate media. That doesn’t mean that we are doomed to fall for deep fake scams. With preparation and vigilance, it will still be possible to identify and avoid even the most compelling, realistic tricks.
Today, the simplest solution is to build processes around the assumption that your organization will become the victim of a deep fake attack. Some steps you can take to reduce the severity of impact include requiring passwords for entry into video calls—an additional barrier to keep unauthorized accounts from joining meetings—and setting up secondary communication channels for verification of particularly sensitive requests, such as those related to finances or granting new user permissions. If the initial request comes through a video chat or email, for example, secondary verification can be done by phone—just be sure you don’t give any information to the requester, not even the phone number to dial, until you confirm the request is legit.
Technology also can be used to help prevent deep fake attacks, as well. In addition to the password requirement in communication applications, using exploit prevention software can stop an attacker from working around the intended function of the software. Another consideration is authenticity verification. This is where technologies like blockchain (and other forms of sender verification like SPF, DKIM and DMARC for emails) come into play. If we can verify that the video, email, or other form of communication we received is authentic and remains unchanged from what was originally sent, then the chances of falling for—or even receiving—a deep fake scam are significantly reduced.
As deep fake technology advances, security professionals need to continue looking for new and innovative ways to prevent these attacks before a potential victim ever has the chance to see them.
Deep fake technology may be nearly 30 years old, but we are only starting to see its true potential. The applications are both exciting and terrifying, but that doesn’t mean we need to live our lives in fear.
If we are aware of the possible uses of this technology, both for good and for evil, we can more easily combat its misuse. Now is the time to begin building strategies that will help to circumvent any potential attacks while also keeping an eye out for existing and emerging technologies that may help prevent such attacks in the first place.
Stay observant, trust your instincts and, above all, verify absolutely everything. Verification will be the most effective means of avoiding deep fake scams as they become more prevalent in the coming years.
What is a cybersecurity vulnerability, how do they happen, and what can organizations do to avoid falling victim? Among the…
We recently hosted a webinar on integrating development and security functions to increase organizational resilience. Industry leaders from Repsol, SAP,…
Authors/Presenters: Xueqiang Wang, Yifan Zhang, XiaoFeng Wang, Yan Jia, Luyi Xing Many thanks to USENIX for publishing their outstanding USENIX…
Unlock the dynamic interplay between cybersecurity and agility in today’s business landscape. Explore how organizations can fortify their defenses, foster…
Insight #1 AI is clearly becoming a problem, with headlines capturing incidents such as a deepfake audio impersonating a Chief…
Segregation of Duties in Oracle ERP Cloud: A Comprehensive Guide to RemediationControlling Risk: An Approach to Automating the Management ofSegregation…