Persona, the identity verification giant powering OpenAI and LinkedIn, has secured the highest certification grade for deepfake detection under Europe’s first standardized injection attack test. The company’s Android software development kit achieved 'High' certification under the CEN/TS 18099 standard, a benchmark published by the European Committee for Standardization in 2024. This validation comes as fraudsters increasingly bypass physical cameras, feeding AI-generated media directly into verification flows to spoof liveness checks.
The Mechanics of the Attack
Injection attacks represent a paradigm shift in identity fraud, moving beyond physical masks to digital manipulation. Instead of presenting a fake face to a lens, attackers use virtual cameras, emulators, and tampered devices to inject stolen or AI-generated video streams directly into the application. Irfan Faizullabhoy, product manager at Persona, noted that customers are facing more sophisticated deepfakes, face swaps, and replay attacks as verification volumes surge. Traditional liveness checks often fail here because the system perceives the injected feed as a legitimate live capture.
Rigorous Evaluation Standards
The certification process was far from superficial. CLR CERT, the certification body of Cabinet Louis Reynaud Labs, subjected Persona’s software to at least 30 person-days of aggressive testing. The evaluation employed a minimum of three injection methods, including rooted devices, tampered camera services, and custom operating system builds. Furthermore, the test utilized at least 15 distinct media types, ranging from deepfake videos and AI-generated faces to spliced video clips, while simultaneously ensuring that genuine users could still complete verification without friction.
Regulatory Pressure and Future Threats
This certification is strategically timed ahead of the European Union Anti-Money Laundering Regulation, which mandates stricter remote customer identification starting July 10, 2027. Daniel Lee, product lead at Persona, highlighted that while deepfake volume is growing, agentic AI poses an even more alarming trend. He described real-time phishing attacks where bad actors mimic verification flows to trick users into submitting genuine IDs and selfies, which are then injected into legitimate sessions. Persona’s defenses now run automatically in the background, providing reason codes like 'virtual camera detected' to help fraud teams audit decisions.
Key Takeaways
- Persona’s Android SDK earned 'High' certification under CEN/TS 18099, the top grade in Europe’s first injection attack standard.
- The test involved over 30 person-days of evaluation using 3+ injection methods and 15+ media types, including deepfakes.
- EU Anti-Money Laundering Regulation covering remote identity proofing takes effect July 10, 2027, driving demand for such certifications.
- Persona serves major clients like OpenAI, LinkedIn, and Square, having verified roughly 1 billion identities globally.
The Bottom Line
Persona’s certification is less about the tech and more about regulatory arbitrage. With the EU AML rules locking in 2027, this 'High' grade gives Persona a defensible moat against competitors who are still treating injection attacks as an edge case rather than a primary vector for identity fraud.