“I could spot an AI deepfake easily.” This seems to be the attitude most people have towards artificial intelligence (AI) – )-generated videos and images. However, a warning has been issued from the science-based solutions for biometric identity verification firm, iProov, as it reveals just 0.1 per cent of 2,000 participants in its latest test were able to successfully identify ake content.
AI-generated videos and images are often created with the purpose of impersonating people, and based on the findings from iProov, there is a large chance impersonators would be successful. Both US and UK consumers were tested, being told to examine a variety of deepfake content, including images and videos. iProov notes that in this test scenario, participants were told to look out for potential use of AI, but in the real world, when consumers are unsuspecting, the success rates of impersonation would likely be even higher.
“Just 0.1 per cent of people could accurately identify the deepfakes, underlining how vulnerable both organisations and consumers are to the threat of identity fraud in the age of deepfakes,” says Andrew Bud, founder and CEO of iProov. “And even when people do suspect a deepfake, our research tells us that the vast majority of people take no action at all.
“Criminals are exploiting consumers’ inability to distinguish real from fake imagery, putting our personal information and financial security at risk. It’s down to technology companies to protect their customers by implementing robust security measures. Using facial biometrics with liveness provides a trustworthy authentication factor and prioritises both security and individual control, ensuring that organisations and users can keep pace and remain protected from these evolving threats.”
An imminent threat
Deepfakes pose an overwhelming threat in today’s digital landscape and have evolved at an alarming rate over the past 12 months. iProov’s 2024 Threat Intelligence Report highlighted an increase of 704 per cent increase in face swaps (a type of deepfake) alone. Their ability to convincingly impersonate individuals makes them a powerful tool for cybercriminals to gain unauthorised access to accounts and sensitive data.
Deepfakes can also be used to create synthetic identities for fraudulent purposes, such as opening fake accounts or applying for loans. This poses a significant challenge to the ability of humans to discern truth from falsehood and has wide-ranging implications for security, trust, and the spread of misinformation.
iProov test findings
When identifying which age ranges were most susceptible to deepfakes, 30 per cent of 55-64-year-olds were tricked by the content. Thirty-nine per cent of those aged over 65 hadn’t even heard of deepfakes. While this highlights the significant knowledge gap on the latest tech and how vulnerable this age group is to this growing threat, they weren’t the only ones unaware of deepfakes. Twenty-two per cent of all those partaking in the test had never even heard of deepfakes before the study.
Despite their poor performance, people remained overly confident in their deepfake detection skills at over 60 per cent, regardless of whether their answers were correct. This was particularly so in young adults (18-34). While it is concerning that older generations have not heard of the technology, it is equally alarming that so many young people have such a false sense of security.
When identifying the deepfakes, 36 per cent of participants struggled more with videos than images. This vulnerability raises serious concerns about the potential for video-based fraud, such as impersonation on video calls or in scenarios where video verification is used for identity verification.
A question of trust
Social media platforms are seen as breeding grounds for deepfakes with Meta (49 per cent) and TikTok (47 per cent) seen as the most prevalent locations for deepfakes to be found online. This, in turn, has led to reduced trust in online information and media— 49 per cent trust social media less after learning about deepfakes. Just one in five would report a suspected deepfake to social media platforms.
Additionally, three in four people (74 per cent) worry about the societal impact of deepfakes, with ‘fake news’ and misinformation being the top concern (68 per cent). This fear is particularly pronounced among older generations, with up to 82 per cent of those aged 55+ expressing anxieties about the spread of false information.
Less than a third of people (29 per cent) take no action when encountering a suspected deepfake which is most likely driven by 48 per cent saying they don’t know how to report deepfakes, while a quarter don’t care if they see a suspected deepfake.
Additionally, most consumers fail to actively verify the authenticity of information online, increasing their vulnerability to deepfakes. Despite the rising threat of misinformation, just one in four search for alternative information sources if they suspect a deepfake. Only 11 per cent of people critically analyse the source and context of information to determine if it’s a deepfake, meaning a vast majority are highly susceptible to deception and the spread of false narratives.
Not all hope should be lost
With deepfakes becoming increasingly sophisticated, humans alone can no longer reliably distinguish real from fake and instead need to rely on technology to detect them.
To combat the rising threat of deepfakes, organisations should look to adopt solutions that use advanced biometric technology with liveness detection, which verifies that an individual is the right person, a real person, and is authenticating right now. These solutions should include ongoing threat detection and continuous improvement of security measures to stay ahead of evolving deepfake techniques.
There must also be greater collaboration between technology providers, platforms, and policymakers to develop solutions that mitigate the risks posed by deepfakes.
Professor Edgar Whitley, a digital identity expert at the London School of Economics and Political Science adds: “Security experts have been warning of the threats posed by deepfakes for individuals and organisations alike for some time. 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.”
