How Federal Benchmarking Exposed Real Separation in Identity Verification

For years, a convenient narrative has shaped the identity verification market: document verification is commoditized.

The logic seemed straightforward. Vendors published similar accuracy claims. Performance percentages appeared tightly clustered. To many buyers, solutions looked interchangeable. That narrative is now under duress.

The U.S. Department of Homeland Security’s Remote Identity Validation Rally (RIVR) delivered something our industry rarely sees: independent, controlled, federal testing of document verification systems under realistic capture conditions. Seven leading vendors were evaluated against genuine government-issued IDs and sophisticated fraudulent documents.

The results revealed a clear and measurable separation in performance.

Six of the seven participating vendors failed to meet DHS “high performing” benchmarks in at least one critical category. Some systems struggled to reliably detect fraud. Others rejected too many legitimate users. Some demonstrated reliability gaps that would create operational instability in production environments.

Microblink, evaluated anonymously as DVS6, was the only system to meet DHS “high performing” thresholds across every measured metric. 

RIVR evaluated more than 2,000 genuine identity documents from 23 U.S. states and Washington, DC, alongside thousands of fraudulent documents. Images were captured across multiple smartphone platforms to simulate real-world user conditions. All materials were validated using government multi-spectral authentication tools.

The takeaway is that document verification is not interchangeable infrastructure. Decision quality varies meaningfully across systems. And in production environments, those differences compound into revenue impact, operational cost, and regulatory exposure.

Reliability Is Now a Business Imperative

Identity verification has matured beyond feature comparison. Enterprises are no longer evaluating demos. They are running global digital onboarding flows, authenticating high-value transactions, and operating under intensifying regulatory oversight.

In production, small performance gaps compound quickly:

• Fraud loss increases when false accepts rise
  
• Revenue declines when legitimate users are rejected
  
• Manual review costs scale when systems fail to resolve ambiguity
  
• Regulatory risk grows when controls behave inconsistently
  

RIVR tested three dimensions simultaneously: fraud detection, genuine user approval, and system reliability. Most systems optimize for one and trade off the others. Balanced performance is significantly harder.

According to the DHS results, Microblink achieved:

• 0.00% System Error Rate
  
• 3.15% Document False Reject Rate
  
• 0.88% Document False Accept Rate
  

DHS established 10% as the threshold for “high performing” systems. Microblink was the only vendor to remain below that benchmark across all three dimensions.

Built for Adversarial Conditions

At Microblink, we have made a deliberate choice to build for adversarial environments, not ideal lab inputs.

Our vertically integrated computer vision models, document authenticity systems, and synthetic fraud defenses are trained using large-scale internal fraud simulation. Our Fraud Lab generates more than 100,000 synthetic images every month to stress-test performance across demographics, devices, and capture variability.

The goal is not to optimize for a single metric. It is to maintain balanced performance under pressure.

Independent benchmarking reinforces an industry truth that is becoming unavoidable: enterprises do not have to accept trade-offs between fraud prevention, customer experience, and operational stability. Balanced decisioning is achievable.

The Industry’s Inflection Point

Accuracy claims alone are insufficient. Enterprises need systems that deliver consistent, high-quality decisions under real-world conditions. Independent validation frameworks like RIVR will play an increasingly important role in separating systems designed for production from those optimized for marketing narratives.

As fraud tactics continue to evolve, especially with generative AI, the defining question for every enterprise will be simple:

Can your identity system make the right decision consistently, at scale, without creating new risk?

Independent federal testing suggests the answer does not look the same across the market. And that distinction now matters more than ever.