Onfido, the global automated identity verification provider, today announced the findings of its 2024 Identity Fraud Report and unveiled its Fraud Lab. The report points to the increasing availability of simple-to-use AI tools as the reason for a 31X increase in deepfakes and a 5X increase in digitally forged identities, making it easier for criminals to carry out sophisticated attacks. Onfido’s Fraud Lab has been launched as a response to these trends. With the ability to mass-produce synthetic attacks, it delivers faster, more accurate AI-powered fraud protection to its customers through its Real Identity Platform, saving them over $3.9B in losses.

The fifth annual Identity Fraud Report highlights the increasing role of AI in perpetrating fraud, as fraudsters turn to online tools and generative AI, such as deepfake apps. Businesses are countering deepfakes by leveraging biometric verification powered by deep learning to detect the presence of a real person, which remains an effective deterrent, creating an AI vs AI showdown. This year, ‘easy’ or less sophisticated fraud also continued to rise, up 7.4% from 2022 and accounting for 80.3% of all attacks. This indicates a persistent trend of quantity over quality, as fraudsters look for minimum effort and maximum reward by launching multiple attacks simultaneously – enabled by broadening access to automated AI-powered tools.

“The confluence of cheap, genuine identity documents appearing on the Dark Web, combined with simpler and free generative AI tools, has created a gangster's paradise,” said Simon Horswell, Senior Fraud Specialist at Onfido. “Creating large-scale, convincing attacks at the click of a button has never been easier for malicious actors, who increasingly use AI to attack en masse. It's become an AI vs AI battleground, where only defensive AI that’s trained on the latest attack vectors at scale can keep up with the evolving threat landscape.”

In the space of just six months, Onfido’s Fraud Lab has helped improve its fraud detection rates on documents by 5X and on biometrics by 9X. Its team of experts, including former police and border agents, collect and label data while measuring fraud detection performance against multiple attack vectors on both document and biometric verification solutions. Training AI models to achieve optimal performance is challenging when only a few new attack incidents are available, so the team dissects the one or two samples they uncover, labeling their key signals. The lab’s unique forensic environment then allows the experts to recreate synthetic derivatives, such as 3D documents or deepfakes in sufficient numbers to generate viable datasets for both training and evaluation. The lab also provides the ability to analyze the anatomy of global fraud rings, enabling Onfido to detect and stop over 267 to date.

“A machine learning model needs around 3,000 fraud samples for effective training. Obtaining these quantities in a short space of time used to be near impossible, which left businesses exposed,” said Vincent Guillevic, Head of Fraud Lab at Onfido. “The Fraud Lab enables real-time data analysis for fraud specialists to identify fraud patterns and emerging trends, while allowing us to create as many as 10,000 synthetic IDs in as little as 30 minutes, so we can now train and benchmark our algorithms quickly and cost-effectively, to provide best-in-class fraud prevention.”

Digital Forgeries and Video Fraud are developing, fast

As the increasing availability of AI tools makes digitally manipulated images even quicker and cheaper to produce, the Identity Fraud Report finds that digital document forgeries jumped 18% in 2023 – with roughly 5X more digital forgeries than observed in 2021, while physical counterfeits declined 17%.

According to Gartner: “Although generative AI is not new, the accessibility of tools using generative AI and the public awareness of its capabilities have accelerated massively.”

Onfido’s Identity Fraud Report also revealed that by far the biggest attack vector used to try to spoof a biometric liveness check is the submission of a video displayed on a screen. This accounts for over 80% of attacks. In fact, video fraud has increased almost 4X year-on-year. This also reinforces the emerging trend of fraudsters turning to techniques like deepfakes or the lesser convincing cheap fakes as a means of attack.

Elsewhere, fraudsters have placed more emphasis on biometric fraud with the average rates twice what they were in 2022. Meanwhile, document fraud overall has leveled out to pre-pandemic levels as bad actors increasingly use a genuine document – obtained on the Dark Web – to bypass the document verification check and rely on AI to switch or swap the face for the biometric scan.

Onfido’s Identity Fraud Report also revealed:

National IDs top criminals hit list: In 2023, fraudsters targeted National ID cards more than any other document, accounting for 46.8% of all document fraud, followed by passports at 26.7%. Not all National ID cards are for international travel and therefore don’t adhere to International Civil Aviation Organization (ICAO) guidelines, making them less robust in terms of protection, and easier targets for fraudsters.

Scalable attacks launched on Financial Services: less sophisticated fraud accounted for 79% of attacks against financial services. By using easy and scalable models fraudsters can determine what sticks before using that loophole to attack en masse. The flood of attacks can also distract organizations from the rarer, but more sophisticated fraud.

Biometrics see markedly lower fraud compared to documents: Biometrics still see a markedly lower rate of fraud compared to document-only checks — which see a 3x higher attack rate. This shows biometrics continue to act as an effective deterrent, particularly solutions with a liveness element.