.svg)
Quick Summary
- Fraudsters in LATAM increasingly use deepfake technology to bypass selfie checks and create synthetic accounts at scale.
- caf needed a modern safeguard flexible enough to work across multiple industries in their orchestration platform.
- By integrating DuckDuckGoose’s DeepDetector, caf gained real-time deepfake screening within their onboarding flows.
- The result: safer onboarding, fewer escalations, and stronger downstream protection for banks, fintechs, and regulated clients.
Introduction
CAF is a leading Brazilian identity verification provider supporting banks, fintechs, marketplaces, and digital platforms across LATAM. Its orchestration layer powers compliant, high-volume onboarding through document checks, facial biometrics, liveness, and fraud prevention. As deepfake-driven attacks accelerated, CAF partnered with DuckDuckGoose to reinforce its KYC pipeline without increasing friction. Together, they deployed an invisible, real-time detection layer that strengthens identity verification while preserving customer experience across diverse industries and risk levels.
The Challenge
CAF needed to address emerging fraud patterns without slowing down user onboarding:
- Deepfakes slipping past standard liveness checks
- Increased manual review load for fraud and AML teams
- High-risk segments (e-commerce, gambling) requiring cost-efficient controls
- Pressure from regulated clients to demonstrate stronger proof of identity assurance
CAF sought an additional fraud layer that would quietly intercept synthetic identities without introducing friction for genuine users.
The Solution
DuckDuckGoose integrated directly into CAF’s selfie capture step through a lightweight API, analyzing the same image used for liveness and returning an immediate pass/flag verdict. The model slots naturally into CAF’s orchestration system, requiring no changes to user experience. Thresholds are fully tunable by customer segment, enabling CAF to adapt fraud defenses based on risk appetite. Optional explainability supports internal fraud teams, while sub-second latency ensures real-time performance at scale. The solution complements CAF’s existing biometrics, closing the deepfake gap without adding cost or friction.
The Result
- 2-4 Million identity checks screened monthly
- 300,000+ synthetic media attempts detected and stopped at onboarding
- 0.2-0.5% false rejection rate across all segments
- 600% reduction in manual reviews and escalations
- 78% reduced dependency on more expensive traditional biometric checks that cannot keep up with deepfake generation development
- 100% strengthened regulatory assurance for CAF’s clients
Business Impact:
CAF reduced costly escalations, strengthened trust for regulated clients, and added a precise, low-friction layer that protects downstream AML teams and product lines.
Conclusion
CAF successfully reinforced its KYC pipeline by embedding DuckDuckGoose at the point of highest risk — the selfie step. The integration delivers precise real-time deepfake detection while maintaining seamless user journeys and operational efficiency across diverse customer segments. This partnership positions CAF to stay ahead of rapidly evolving synthetic media threats and continuously strengthen identity assurance for its clients.
Key Takeaways
- Deepfake-enabled fraud is accelerating across LATAM; banks and IDV providers require AI-driven defences that adapt in real-time.
- caf strengthened its platform by adding a layer that complements existing biometrics without impacting user experience.
- Deepfake detection is now a required control for modern KYC flows
- Layered identity defenses must adapt by segment and risk profile
- Invisible, real-time AI protection prevents fraud without harming conversion
- DDG’s DeepDetector helps regulated clients stay ahead of evolving threats while maintaining seamless onboarding.
