Bad actors operate across banks, fintechs, and payment corridors. Chefoba lets institutions train a shared detection model on that cross-institutional pattern, without any one of them seeing another's customer data.
Join the waitlistThe problem
A single fraud ring or laundering network typically touches multiple institutions. Each one sees only its own slice of the activity, which is often not enough to flag it as suspicious on its own. Pooling data to close that gap is usually a non-starter: competitive sensitivity, data protection law, and contractual restrictions all stand in the way.
How Chefoba addresses it
Each institution trains locally on its own transaction and account data. Chefoba aggregates the resulting model updates into a detection model that reflects patterns across every participant.
Institutions with stronger, more consistent data contribute more heavily to the shared model, tracked automatically through Chefoba's trust scoring.
Every aggregation round produces evidence: which institutions contributed, under what policy, with what outcome, ready for your compliance team and your regulator.
Why this matters for accuracy, not just privacy
In our evaluation, subgroup auditing by payment corridor found a fairness gap of 0.797 in true positive rate between corridors, a gap that was completely invisible when only looking at aggregate model accuracy. A model that looks strong overall can still be systematically worse for specific corridors, customer segments, or institution types. Chefoba's auditing tools are built to surface exactly this kind of gap before deployment, not after a regulator or a customer finds it.
Read the fairness findings →We're onboarding a limited group of institutions ahead of general availability.
Join the waitlist