You’ve just come home from a long day of work and are about to eat dinner when suddenly your phone starts buzzing. On the other side is a loved one, perhaps a parent, child, or childhood friend, begging you to send them money immediately.
You ask them questions and try to understand them. There’s something wrong with their answers, which are either vague or uncharacteristic, and sometimes there’s a curious delay, almost as if they were thinking a little too slowly. Still, you’re sure it’s definitely your loved one speaking: that’s their voice you hear, and the caller ID shows their number. Underlining the strangeness of their panic, you dutifully send the money to the bank account they provide you.
The next day you call them back to ask if everything is okay. Your lover has no idea what you are talking about. That’s because they never called you – you’ve been duped by technology: a voice deepfake. Thousands of people were Scammed this way in 2022.
If computer security researcherswe see that the continuous advancements in deep learning algorithms, audio editing and engineering, and synthetic voice generation have made it increasingly possible to convincingly simulate a person’s voice.
Even worse, chatbots like ChatGPT are starting to generate realistic scripts with adaptive real-time responses. By means of combining these technologies with speech generationa deepfake goes from a static shot to a live, lifelike avatar that can convincingly hold a phone conversation.
Clone a voice
Creating an engaging, high-quality deepfake, be it video or audio, isn’t the easiest thing to do. It requires a wealth of artistic and technical skills, powerful hardware, and a fairly hefty sample of the target voice.
A growing number of services are offered produce moderate to high quality voice clones for a feeand some voice deepfake tools need preview of just for a minuteor even just a few seconds, to produce a voice clone that could be convincing enough to fool anyone. To convince a loved one – to use in an imitation scam, for example – will likely require a significantly larger sample size.
Protection against scams and misinformation
That said, we at the DeFake project from the Rochester Institute of Technology, the University of Mississippi and Michigan State University, and other researchers are working hard to detect video and audio deepfakes and limit the damage they cause. There are also simple and everyday actions you can take to protect yourself.
For starters, voice phishing, or “vishing”, as described above, are the most likely voice deepfakes to encounter in everyday life, whether at work or at home. In 2019, a energy company was scammed for $243,000 when criminals simulated the voice of the parent company’s boss to order an employee to transfer money to a supplier. In 2022, humans were scammed for an estimated $11 million through simulated voices, including from close, personal connections.
What can you do?
Be prepared for unexpected calls, even from people you know well. This isn’t to say you should schedule every call, but it helps to at least send an email or text. Don’t rely on caller ID either, since that can also be faked. For example, if you get a call from someone claiming to represent your bank, hang up and call the bank directly to confirm the legitimacy of the call. Make sure to use the number you wrote down, saved in your contacts list or can find on Google.
In addition, be careful about your personally identifiable information, such as your social security number, home address, date of birth, phone number, middle name, and even the names of your children and pets. Scammers can use this information to impersonate you to banks, brokers and others, enriching themselves while bankrupting you or destroying your credit.
Here’s another piece of advice: know yourself. In particular, know your intellectual and emotional biases and vulnerabilities. This is generally good life advice, but it is essential to protect yourself from manipulation. Scammers usually try to find out and then exploit your financial concerns, your political affiliations or other inclinations, whatever they may be.
This alertness is also a decent defense against disinformation with voice deepfakes. Deepfakes can be used to take advantage of your confirmation bias, or what you tend to believe about someone.
If you hear an important person, whether from your community or the government, say something that either seems very unusual to them or confirms your worst suspicions about them, it is wise to be wary.
