New · DBX Suite + AI is here — reimagine digital banking.New · DBX Suite + AI is here. Explore DBX →

Cross-border payments, reimagined with CrossPay Exchange.Meet CrossPay Exchange. Learn more →

P2P Connect — AI-powered peer-to-peer lending.AI-powered P2P lending. Discover P2P →

See DBX in action — book a live demo with our team.See DBX in action. Book a demo →

All articles
Risk & ComplianceBlog

Fraud in Instant Payments: Detecting the Undetectable

Instant payments are irrevocable and settle in seconds. That is wonderful for customers and a gift to fraudsters, because the overnight batch review that used to catch bad transactions no longer has time to run.

DODaniel OkoroPrincipal Solutions Architect, OmnicoreJun 8, 2026 · 7 min read · 2.9k reads

The 400-millisecond problem

When a payment clears in under a second, fraud decisioning has to happen in-line, before authorisation. That means real-time feature computation, a model that scores in milliseconds, and a policy engine that can hold, step-up or decline without adding perceptible latency.

Behaviour beats blocklists

Static rules and blocklists cannot keep up with authorised push-payment scams, where the customer themselves is manipulated into paying. Behavioural signals — device, session, payee history and velocity — catch the anomalies that a rule never anticipated.

Design for recovery too

No model catches everything. Fast confirmation-of-payee, transaction limits for new payees and a rapid recall workflow limit the damage when something does slip through.

Key takeaways
  • Score fraud in-line, before authorisation, in milliseconds
  • Use behavioural signals to catch scams rules miss
  • Layer confirmation-of-payee and recall to contain losses

See it in your own environment

Request a tailored demo of the products behind this research.

Request a demo

Keep reading