Fraud detection built where the fraud actually happens.

RiskGuard exists because the fraud patterns in West African mobile money look nothing like credit card fraud in the US. SIM swap attacks, agent collusion rings, split transactions designed to stay below reporting thresholds — these are specific, local patterns that global fraud tools don't catch.

We built RiskGuard at JoonaPay after spending months trying to make generic fraud detection tools work for UEMOA mobile money. The rule engines couldn't express “flag if 3+ distinct users cash out at the same agent within 30 minutes after midnight.” The ML models were trained on Western card fraud and missed everything.

So we built our own. A YAML-based rule engine for speed, Neo4j for graph pattern detection, and sanctions screening tuned for French/Arabic name transliteration. We run it on every transaction in the Korido wallet — and now it's available as a standalone API.

The team

RiskGuard is built by the engineering team at JoonaPay, based in Abidjan, Côte d'Ivoire. We've built payment systems processing transactions across 8 UEMOA countries. We know the regulatory environment (BCEAO), the mobile money ecosystem (Orange, MTN, Wave, Moov), and the fraud patterns that come with it.

Open source commitment

The rule engine and evaluation API will be open-sourced under MIT license. The graph analysis module (Neo4j integration) is available in the Growth and Enterprise plans. We believe fraud detection infrastructure should be transparent — you should be able to audit exactly how decisions are made.