- Deepfakes increasingly trick users into fraud and scams
- AI-driven attacks now involved in 16% of breaches
- Spotting fakes requires watching facial/audio glitches and urgent money requests
Deepfakes, the technology that uses artificial intelligence (AI) to synthesize realistic audio, video, and images, of real people, has improved to the point where even attentive people can be tricked with relative ease.
By creating multimedia showing celebrities, politicians, or tech leaders, saying things they didn’t say, or doing things they never did, internet scammers managed to get people to authorize fraudulent transfers, or to get innocent people to “invest” into fake projects on fake investment platforms. In some cases, they even managed to create a fake emergency (such as a car accident, or a hijacking attempt), that made family members rush into making payments.
These days, 16% of all breaches involve AI-driven attacks (as per IBM data), with phishing and deepfakes being among the most commonly cited methods, Danny Mitchell, cybersecurity writer at Heimdal Security, claims - and after drilling deep into the rising scam technique, he discovered that there are still ways to spot a deepfake, and has shared his findings with TechRadar Pro.
Four actionable steps to spot a deepfake
According to Mitchell, one of the ways to spot a deepfake is to watch for unnatural facial movements or blinking.
“Deepfake videos can struggle with the subtle mechanics of human expression,” Mitchell writes. “Watch for blurred facial edges, inconsistent blinking, or expressions that don't match the emotion being conveyed.”
He also argues that audio can sometimes be slightly off in deepfake videos. AI-generated voices can carry a faint flatness, or an unusual rhythm, while background noise may sound artificial.
Also, deepfake videos often have mismatched lip movements, as synchronization between speech and lips is often imperfect. This is particularly visible at faster talking speeds.
Therefore, spotting deepfakes comes down to pixel hunting. But there are other ways to identify a fake, and that is to be mindful of the content of the message. Mitchell says that urgent requests for money or sensitive information, pressuring users to act quickly, is the usual red flag that users should pay attention to.
