Full-cycle verification platform, Sumsub, has launched a deepfake detection feature integrated into its video identification solution.

This comes as artificial intelligence (AI)-powered fraud increasingly targets businesses, not just individual users. 

Sumsub’s 2023 Identity Fraud Report revealed a tenfold increase in the number of deepfakes detected globally across all industries from 2022 to 2023, with crypto and fintech jointly constituting 96% of these cases.

Deepfake incidents in the fintech sector increased by 700% in 2023 compared to the previous year, underscoring the importance of ensuring malicious actors are screened out during verification processes.

Sumsub says that its deepfake detection feature represents a step forward in combating AI-driven identity fraud, particularly as fraud tactics become increasingly sophisticated.

Deepfake detection years in the making

“AI technology and its risks aren’t new and we’ve spent many years developing deepfake detection tools long before they were thrust into the mainstream,” said Andrew Novoselsky, Chief Product Officer, Sumsub.

“As AI technologies advance, so do the tools available to fraudsters.

“Unfortunately, cases like these will continue to occur, making our first-of-its-kind solution that embeds deepfake detection into real-time video identification so imperative for the security of businesses and consumers alike.”

In the last year, Sumsub carried out over a million video identity checks for firms, securely verifying their clients’ end-users.

Along the way, Sumsub’s AI and machine learning team noticed a pattern in video verification interviews being susceptible to deepfake attacks.

This follows the launch of Sumsub’s deepfake detection solution, embedded in its in-house Liveness product.

Unlike methods that focus on detecting deepfakes within static images or recorded videos, Sumsub’s new solution operates in real-time during video interviews, further securing the identification process.

The deepfake detection technology is designed to identify and thwart fraudsters attempting to manipulate real-time video interviews for malicious purposes.

Video identity verification is mandatory in a number of jurisdictions, including Germany, Switzerland, Austria and Estonia.

However, it can also be used as a reputable last resort for user verification in most markets, when businesses want to assure an additional layer of security during onboarding.

Learning to trust your eye

“We’re frequently seeing increasing demand from our clients to carry out video interviews, so we understand the importance of ensuring these are completely secure,” said Pavel Goldman-Kalaydin, Head of AI & ML, Sumsub.

“Video identification is often viewed as the ultimate defense against fraud during onboarding, but as deepfake technology evolves, trusting our eyes during live video is no longer foolproof.

“To address this, we’re actively enhancing our detection capabilities.

“As deepfakes continue to impact business operations, as well as everyday individuals, we are committed to working continuously to expand our detection capabilities to protect firms and their clients.”