Fraud Prevention using Machine Learning – Transaction Risk Monitoring Qorus Reinvention Awards - APAC 2024-2025 - Winner

Real Time ML Fraud is an advanced machine learning-powered fraud scoring engine. It detects fraud by scoring each card transaction of 6 million transactions daily in real-time. Using historical data, it analyzes cardholder data, transaction patterns, and channel attributes to generate real-time fraud likelihood scores. These scores enhance capturing more fraudulent transactions and reducing false positives while handling complex fraud schemes across POS, ATM, & CNP for both credit and debit cards. It improves customer experience and increases bank savings. The innovation in ML Fraud lies in its ability to explain complex fraud concepts through data-driven approach & modelling fraud as a function of user, channel, product details, & their derivatives. This allows ML Fraud to capture subtle & evolving patterns in fraud that traditional systems miss, ensuring continuous improvement in fraud detection accuracy. Summarising the benefits of the above innovation: 1. Since Inception: Achieved Split second response viz. 15 milliseconds (ms) for scoring the transaction; Capacitized for 1000 TPS with response time within 15 ms; Simplified user adoption. 2. Within 6 months of implementation: Real time fraud catch incremented by 30%; Reduced 30% overall false positives; Gained Customer Trust; 3x ROI achieved within 6 months.