This post is also available in:
Hebrew (Hebrew)
There have been numerous reports that voice messaging is becoming increasingly popular, especially among younger generations, and bad actors are trying to take advantage of this trend.
AI-generated audio deepfakes are already a known concept and are increasingly being used in cyber attacks, especially those involving personal information. Another common attack type is the “account takeover attack”, which is the most valuable attack for bad actors. Aaron Painter, CEO of security solutions provider Nametag, explains: “When you take over someone’s account, you control everything that account has access to. For employee accounts, this means planting ransomware or accessing sensitive company data. For customer accounts, you can take over social media and bank accounts. Deepfakes make it easy to do that. Deepfakes are a superpower for bad actors.”
According to CyberNews, the main reason audio deepfakes are on the rise is that they are now easier to create: Whereas in 2020 it took 20 minutes of audio to create an audio deepfake, now it only takes a few seconds. Painter added that higher quality fakes may require more time, but attacks can still be successful with relatively low quality deepfakes.
Jason Glassberg of Casaba Security predicts that the next wave of cyberattacks will exploit the younger generation’s habit of sending voice notes. “Most people, especially younger people, are well aware of the risks of phishing, smishing, and even conversation hijacking in text-based interactions,” he says. “They tend to be skeptical if there’s something odd or not right about a written message, such as a request for money. But a voice message is different. A voice message is more persuasive.”
When determining whether audio is a deepfake, Glasberg believes the best approach is to evaluate the context and consider who the sender is and what channel is being used (for example, an audio message in a large public group chat is more suspicious than one in a private one-on-one chat).
They also say to look out for edits, unnatural noises, and speech that doesn’t match the speaker’s personality, and recommend paying attention to breathing, as many deepfake voices do not.