Most fraud detection systems evaluate transactions one rule at a time, across separate queries and services. In this ODBMS interview, Volt Active Data Senior Systems Software Engineer Marina Popova explains why that approach breaks down at scale and what a genuinely sub-millisecond fraud detection architecture actually requires. She walks through atomic rule evaluation using VoltDB stored procedures, how TIME_WINDOW materialized views make velocity-based fraud rules viable at high throughput, and how adding spatial clustering with MaxMind GeoIP2 surfaces coordinated attack patterns that per-account logic simply cannot see.
Featured Resources
