FraudGuard™ screens a single payment, application, or transaction the moment it appears — before it joins the graph, before a dollar moves. It pattern-matches each record against known fraud signatures, threshold breaches, and anomalies versus expected program behavior. A hit doesn't get recovered after the fact. It's held at the decision.
Built on Decision Governance™ by DataVisuals™ · Fraud Governance · screened at the decision
FraudGuard works on the record in front of it — one payment, one application, one transaction. It doesn't need the chain or the entity history to raise its hand. Each record gets three reads.
Matches the record against a library of known fraud patterns — structures, sequences, and shapes that have signalled fraud before.
Tests the record against program limits and ceilings. An amount, a frequency, or a timing that crosses a hard boundary gets flagged on contact.
Compares the record to expected behavior for its program. What's statistically off — not just over a fixed line — surfaces as a signal for review.
Decision Pipelines™ runs three independent signal sources, each asking a different question of the same federal dollar. They don't replace each other — they catch different things. FraudGuard is the single-record lens.
Single record · FraudGuard™Signatures, thresholds, and anomalies on an individual payment, application, or transaction — no context required.
Resolves records from many sources to canonical tokens the graph hangs off — and raises its own alerts when identities collide.
Conservation, cycles, skip-tiers, orphans — anomalies in the shape of the path that only appear end to end.
See the graph →Before the flow graph can connect anything, every record has to resolve to a canonical entity. Identity resolution does that — matching across SAM, SSA, IRS, and state eligibility databases. It's a prerequisite for the graph, and a signal source in its own right.
One strong key — an SSN, a UEI — pointing at divergent attributes. Or one identity reached through mutually inconsistent strong keys. The system refuses to silently merge them.
Confidence in the middle band — too strong to drop, too weak to trust. Rather than auto-merge or discard, the match is promoted to a governed Decision for human adjudication.
FraudGuard is designed to screen each record against the authoritative federal sources — so duplicate identities, excluded parties, and deceased recipients are caught at the decision, not discovered in a post-payment clawback.
The screening logic and the governance around it are built. The live program feeds that put real records through it are the integration step ahead — so the working prototype runs on seeded, synthetic records.
FraudGuard catches the single record. Identity resolution catches the entity. The graph catches the chain. Every hit lands in the same place — held on the record, with rationale, before the money moves.
Screen at the decision. Not in a post-payment clawback.
