Preventing Fraud @ Grab
summary
Systems at Grab perform real-time detect and take action on payments using device fingerprints, rooted-device checks and behavioral features aggregated per user and merchant. A streaming pipeline evaluates ~15 million transactions and uses rolling 30-minute window aggregates and time-series models to surface anomalies. The decision path targets fraud within 750 ms decision latency, enriches events with device and transaction history, and triggers blocks, velocity limits or manual review; implementation notes cover rooted devices, feature sets, aggregation keys and thresholds proven in production.
AI-written summary · about
chapters (7)
- 0:00Introduction and welcome
- 7:16Fraud detection systems
- 14:28Real-time transaction monitoring
- 22:08Biometric verification methods
- 28:03Rolling window aggregation
- 34:07System performance and challenges
- 36:57Q&A session
related talks
13mBehavioral Analytics for Real-Time Fraud Prevention
10mReal-Time Fraud Prevention at Scale
16mAI-Driven Behavioral Analytics for Real-Time Fraud Prevention
28mFortifying Digital Trust: Next-Gen Strategies in Fraud Prevention
18mReal-time Fraud Detection using Deep Learning
45mModern fraud trends in RTB
37mFraud prevention in the age of online privacy
26mAI-Powered Fraud Detection in Cloud-Native Digital Payments