Case study · Financial services
PIX fraud prevention in 50ms. 25 million transactions/day
The fraud prevention program for PIX — Brazil's instant payment system — at CAIXA: detecting and preventing fraud in financial transactions with AI, patented artificial neural networks and real-time scoring in under 50 milliseconds.
Volume and speed that redefine fraud prevention
25M+
Daily transactions in real time
<50ms
Scoring time across the PIX ecosystem
360°
View of the fraud prevention process
ML
Continuous probabilistic models
What we did
A transactional fraud prevention platform in real time
- Real-time scoring in under 50 ms for fraud across the PIX ecosystem and other transactions.
- Behavioral models built on patented artificial neural networks, embedded in the solution's engine.
- Empowering business teams building proactive rules and managing restricted-party lists directly in the platform.
- Transactional data replication for analytics and data science.
Workstreams
Four projects in production
PIX fraud prevention
Real-time scoring for PIX, covering monetary operations (withdrawal, cash back, refund) and non-monetary ones (registering and deleting payment keys).
Other financial transactions
Withdrawals, transfers, deposits, payments, benefits, lottery-agent operations and debit cards.
Non-monetary transactions
Card issuance, device changes and registrations — enriched with behavioral data.
Analytics & investigation environment
Self-service analytics, dashboards and ML on near-real-time data for investigation.
Benefits
More security in every transaction
- Efficient, secure detection of suspicious financial transactions
- Lower financial and operational losses from fraud
- Real-time transaction scoring, integrated into the authentication flow
- Flexible rule changes by fraud strategy teams
- ML-driven detection based on probabilistic models
- A 360-degree view of strategic fraud prevention data
- Regulatory compliance
Transactional fraud prevention in milliseconds
Explore Vert's solutions for financial services — MAIN and Analytics applied to very-high-volume operations.