Fraudsters are now feeding deepfake videos straight into selfie ID checks, skipping the phone camera entirely, and Persona, the identity company that verifies users for OpenAI and LinkedIn, just earned the top grade in Europe's first test for stopping them.
Key Takeaways
- Persona’s Android software development kit earned CEN/TS 18099 High certification for injection attack detection
- High is the top level in the standard published in 2024 by the European Committee for Standardization
- CLR CERT tested the software for at least 30 person-days using multiple attack methods and media types
- The European Union Anti-Money Laundering Regulation covering remote customer identification applies from July 10, 2027
Persona announced that its Android software development kit earned High certification under the European Committee for Standardization’s CEN/TS 18099 injection attack detection standard.
The evaluation was run by CLR CERT, the certification body of Cabinet Louis Reynaud Labs. Persona said the certified protection is available to all customers and can be activated by working with its team.
How The Test Worked
Injection attacks skip the camera entirely. Instead of holding up a photo or mask to a lens, fraudsters use virtual cameras, emulators, and tampered devices to feed stolen, manipulated, or artificial intelligence-generated media straight into the verification flow.
That matters because the system can see the result as a real live capture, while traditional liveness checks are not built to catch it. Irfan Faizullabhoy, product manager at Persona, said customers are facing more sophisticated deepfakes, face swaps, and replay attacks as identity verification volumes rise.
For the High level, evaluators attacked Persona’s Android software over at least 30 person-days. The test used multiple ways to tamper with the capture process and multiple kinds of fake media, while also checking that real users could still complete verification.
| High-Level Test Requirement | Evaluation Detail |
|---|---|
| Attack Time | At least 30 person-days |
| Injection Methods | At least 3, including rooted devices, tampered camera services, and custom operating system builds |
| Attack Media Types | At least 15, including deepfake videos, AI-generated faces, and spliced video clips |
Persona said its defenses run automatically and in real time. Users take a short guided selfie while the system analyzes video, device, and camera signals in the background. Fraud and risk teams then get reason codes such as "virtual camera detected" to explain decisions during audits and appeals.
Also Read: Sumsub Targets AI Agent Risk With 3 Controls For Banks, Insurers And Marketplaces
Why This Matters For Persona
This is a new benchmark in a growing fraud fight. The European Committee for Standardization published CEN/TS 18099 in 2024 as the first standard for testing injection attack detection, with three grades: Basic, Substantial, and High.
The timing is important because the European Union (EU) is raising the bar for remote identity proofing. The EU Anti-Money Laundering Regulation, which covers remote customer identification, applies from July 10, 2027. European identity-proofing standards already point to this testing too, including European Telecommunications Standards Institute TS 119 461 for trust service providers.
Persona also tied the certification to a change in the threat itself. Daniel Lee, product lead at Persona, said, "Deepfake submissions keep growing in volume and sophistication, but agentic AI is an even more alarming trend." He said Persona is seeing real-time phishing attacks where bad actors mimic verification flows, trick people into submitting government IDs and selfies, and then try to inject that genuine media into legitimate sessions.
What To Watch Next
The immediate next step is customer rollout. Persona said customers can turn on the certified injection attack detection by contacting their account team, which means the feature is available now rather than waiting for a later release.
The certification also comes with outside validation. CLR Labs is accredited under the international standard for testing laboratories, and it edits CEN/TS 18099. Kévin Carta, director of CLR Labs, said the evaluation was built to test defenses against injected video, tampered apps, and modified operating systems.
Persona’s size gives the result broader weight. Founded in 2018 and headquartered in San Francisco, the company has verified roughly 1 billion identities, is available in more than 200 countries and territories and 20 languages, and says its fraud prevention and identity theft protection have stopped more than 300 million fraud attempts. It supports consumer-facing businesses including LinkedIn, OpenAI, Twilio, Square, Udemy, Travelex, and Gusto. Its certified status is also listed on the CLR CERT certified products list, making this a standard to watch as the 2027 rule date gets closer.
Read Next: EU’s AI Act Flags Hiring Software As High Risk, Fines Reach €35M
Originally published on Fathom News.