Socure, the provider of artificial intelligence for digital identity verification, sanction screening, and fraud prevention, launched its next generation Document Verification (DocV) solution, purpose-built “to tackle the rash of stolen identities, spoofing, and highly sophisticated deepfake attacks that are accelerating in their complexity across nearly every segment of the digital economy.”
Open source generative AI tools “have intensified the enablement of easy-to-create fraudulent identities alongside the inexpensively purchased stolen IDs and personally identifiable information (PII) found on the black market.”
Legacy document verification solutions “that aren’t investing extensively in R&D are allowing fraudsters to beat their systems using many of these techniques: tampered headshots, non-live selfie presentation attack, duplicative headshots of the same person across varied PII/documents, mismatch of PII provided in form field vs extracted from document or barcode extraction, substantially different age prediction between headshot and document date of birth, document invalidity, deepfake selfie of true ID owner’s face, forged or blurred barcodes, image from a digital screen, repeat device usage across multiple documents in short periods of time and more.”
Sprinting ahead of these rapidly evolving complexities, many that are “used in conjunction with one another in a single attack, Socure has launched a proprietary solution that combats every single dimension of a potential attack at once, making it prohibitively expensive, time-consuming, and outright unattainable for the fraudsters to evade detection – all in less than 2 seconds and with no impact to the user experience.”
This multi-layered defense approach “ensures that increasingly complex attacks cannot circumvent the system that now analyzes 1000s of document signals in real-time with a broader view of the identity risk by also analyzing PII, barcode data, device and behavioral intelligence, geolocation, and biometric signals, all resulting in the attainment of Socure’s market-leading accuracy and automation.”
In addition, Socure’s DocV alleviates consumer friction concerns “presented by typical document and biometric verification solutions, bringing the verification speed down to a P95 of 1.5 seconds, compared to the industry average of more than 30 seconds, with some requiring minutes to return a decision.”
Johnny Ayers, founder and CEO of Socure said:
“Socure continues to get better and better for our partners with each passing day, improving user experience, accuracy, speed to decision and operational efficiency. The Socure brand is known for continuous innovation as we strive for perfection and this announcement is representative of our continued advances. With generative AI fraud tools in the hands of bad actors, we are seeing attacks at massive scale with the use of incredibly accurate deepfakes and manipulated documents that easily fool legacy biometric and template-based solution providers.”
For example Dave, a Socure neobank customer since 2020, has “seen a dramatic 57% increase in DocV auto approval rates resulting in a 21% lift in their overall acceptance rate.”
Additionally their team “has fully automated the experience eliminating the need for manual reviews. It’s also important to their business to stay ahead of growing generative AI threats.”
In its most recent innovation cycle, the Socure document verification solution “is shutting down fraudsters who are wielding generative AI tools to spin up sophisticated, hyper-realistic deepfakes that fool legacy document verification checks, mimicking both credentials and the selfies taken to match those IDs.”
These deepfakes are not discernible “to the human eye, and they’re outpacing the traditional machine learning models used by most providers in the market.”
Attacking generative AI for ‘bad’ with generative AI for ‘good’, the Socure data science team generated hundreds of thousands of “carefully curated deepfakes using more than 20 different AI generators to ensure production of a very realistic diversity of body shapes, ages, skin tones, postures, lighting conditions, backgrounds and ethnicities.”
These deepfakes train Socure’s selfie models to thwart spoofing attacks with incredible precision.
The deepfake selfie detector is only “the newest layer of defense against these robust attacks. Socure also delivers detection for instances in which fraudsters attempt to submit a photo of another image or capture a digital image off a screen, as well as proprietary injection attack detection (with multiple patents having already been filed), which prevent digitally created deepfake selfies from being injected into the capture or verification session.”