A vanishingly small number of people can correctly identify deepfake content, even when primed to look for it, according to research from biometric security provider iProov.

In a new study which tested 2,000 UK and US consumers, iProov found that just 0.1% could correctly differentiate between all deepfake and real stimuli, including both images and videos, despite being aware of what they were looking for. 

According to iProov, these startling results suggest that in real-world scenarios, where people are less aware, the vulnerability to deepfakes is likely even higher.

Deepfake videos proved to be more challenging to identify than deepfake images, with participants 36% less likely to correctly identify a synthetic video compared to a synthetic image, however, despite their poor performance, 60% of people remained overly confident in their deepfake detection skills , regardless of whether their answers were correct.

Just as concerning is that over a fifth of all participants (22%) said they had never heard of deepfakes before the study, with that figure rising to 30% of 55-64 year olds, and 39% of those aged 65 and over.

These numbers are evidence of a significant knowledge gap among the general public, which could increase susceptibility to this emerging threat.

Last year, another report from iProov highlighted a 704% growth in deepfake face swap attacks, as well as a 255% increase in digital injection attacks – sophisticated, highly scalable, cyber-attacks that can be used to spoof biometric verification and authentication systems using deepfakes.

There has also been a rise in the number of deepfake-as-a-service marketplaces on the dark web, where cyber-criminals can more easily purchase fraud tools allowing them to execute attacks at scale, manipulate metadata, and obscure detection.

The technology is being widely used, especially by scammers hoping to trick vulnerable people with celebrity endorsements. For example, the UK’s Advertising Standards Agency recently warned it had observed an increase in the use of celebrity deepfakes, including one featuring the likeness of King Charles recommending cryptocurrency investment.

“Criminals are exploiting consumers’ inability to distinguish real from fake imagery, putting our personal information and financial security at risk,” said Andrew Bud, founder and CEO of iProov. 

“It’s down to technology companies to protect their customers by implementing robust security measures.”

Thankfully, according to this latest study, the growing prevalence of deepfake technology is beginning to be recognised, and is starting to drive widespread concerns among those who understand the threat.

Three in four people (74%) said they worry about the societal impact of deepfakes, with fake news and misinformation being their top concern (68%), a fear particularly pronounced among older generations, with up to 82% of those aged over 55 expressing anxiety about the spread of false information.

Recommended reading

• 82% of Phishing Toolkits Use Deepfakes
• Comment | The Rise in QR Code Attacks
• NCSC Warns Ransomware Threat to Rise with AI

However, the study found that despite their fears, consumers currently don’t do much to help tackle the problem.

Almost a third of people (29%) take no action when encountering a suspected deepfake, with a quarter saying they did not care if they encountered one online. Just one in five said they would report a suspected deepfake to social media platforms, while around half (48%) said they didn’t know how to go about reporting them.

Only one in four told the study they search for alternative information sources if they suspect a deepfake, and just 11% critically analyse the source and context to determine if what they are dealing with is real, leaving the vast majority highly susceptible to deception and the spread of false narratives.

“Security experts have been warning of the threats posed by deepfakes for individuals and organisations alike for some time,” said Professor Edgar Whitley, a digital identity expert at the London School of Economics.

“This study shows that organisations can no longer rely on human judgment to spot deepfakes and must look to alternative means of authenticating the users of their systems and services.”