Fraud prevention for every business

Stop fraud before it moves. Prove every decision.

A cargo release at the gate, a vendor payment, an insurance or healthcare claim, a refund, a loan, a SIM swap, a benefit payment, a sign-in: Omnareth checks each high-risk action against the record and your rules before it happens, recognizes who tried before, and seals every decision so it can be proven unaltered later.

11 reference scenarios75 ms median decision, lab testSHA-256 sealed records

Decisions · 11 reference scenariosverified on chain
#EventDecision
01 ROAD Impostor driver, real release order DENY FICTITIOUS_PICKUP
02 SEA Ship-to-ship transfer reported DENY SHIP_TO_SHIP_TRANSFER
03 SEA Forged original bill of lading DENY INVALID_BILL_OF_LADING
04 PAY 'Northwind' emails new bank details from northwlnd-steel.com DENY BANK_CHANGE_UNVERIFIED
05 CLAIM A stranger claims on another policy with the same police report DENY CLAIMANT_NOT_INSURED
06 REFUND "Never arrived" on a delivered order, refund to a gift card DENY INR_DELIVERED
07 LOAN Synthetic identity, verification fails DENY KYC_FAILED
08 LOAN Genuine disbursement to the borrower's verified account ALLOW
100/100 decisions across 11 scenarios as expected · every record verified on chain
100/100decisions across 11 reference scenarios, correct and verified on chain
3/3database tampering methods caught in testing
22 µsmedian rule evaluation across 34 cargo rules, lab test
75 msmedian end-to-end decision, fraud-checked, stored and audited, lab test

Performance measured on 4 vCPU Intel Xeon @ 2.1 GHz, Node.js 22, PostgreSQL 16 and a four-validator PACT network on the same host, production settings (sign-in, field encryption).

The problem

The costliest fraud needs no force, only a convincing lie

A driver who was never dispatched, a bill of lading that was never issued, an email from a vendor that is not the vendor. Each succeeds because the person deciding cannot see the whole record, because the same fraudster is never recognized twice, and because nobody can later prove what the record said.

Fictitious pickup

A driver quotes the real release order number and drives away with the container.

Forged bill of lading

A copied or invented bill is presented at the port of discharge and the cargo is released to the wrong party.

Payment diversion

An email from a lookalike domain changes a vendor's bank details, and the next invoice is paid to a fraudster.

Rewritten records

When a loss is investigated, the log sits in databases the people involved can edit. Nobody can show it is the log as it was.

Solutions

Built for where fraud is most expensive

Each solution ships with reference rules and a scenario that runs end to end, with every decision verified. The same platform takes your own process: define the record, the events and the rules.

Cargo by road and sea

Custody handoffs, vessels and releases: impostor drivers, forged release orders and bills of lading, seal swaps, diversion at sea.

Cargo by road and sea

Payments

Vendor bank changes and invoice payments: business email compromise, payment diversion, duplicate invoices, self-approval.

Payments

Insurance claims

Claims and payouts: claimants not insured, copied police reports, pre-inception losses, duplicate claims, redirected payouts.

Insurance claims

E-commerce refunds

Refund requests: false "not received" claims on delivered orders, reused return labels, refunds to other destinations.

E-commerce refunds

Lending

Applications and disbursements: synthetic and stolen identities, shell employers, loan stacking, disbursement to others' accounts.

Lending

Healthcare claims

Provider claims: services not rendered, copied prior authorizations, duplicate billing, excluded providers, possible upcoding.

Healthcare claims

Telecom

SIM swaps, port-outs and device orders: account takeover, unauthorized ports, financed phones to drop addresses.

Telecom

Government benefits

Benefit claims and payment changes: stolen and deceased identities, redirected payments, weeks claimed twice.

Government benefits

Cybersecurity

Sign-ins, security changes and privileged actions: credential stuffing, MFA push bombing, account takeover, insider data exports.

Cybersecurity

Identity verification

Document and selfie checks at onboarding: forged documents, deepfakes, presentation attacks, stolen and synthetic identities.

Identity verification

How it works

Beside the systems you already run

Omnareth does not replace your transport, terminal, ERP or payment systems. It gives all of them the same rules, the same memory of who tried, and the same tamper-evident record, through one API.

Step 1

Send the event

A gate system, a vessel feed, a tracker or your payment run sends each high-risk action to Omnareth before it happens.

Step 2

Get a decision

Omnareth checks it against the record, your rules and what it knows about everyone involved, and returns ALLOW FLAG or DENY with reasons.

Step 3

Prove the record

Every decision is stored with a SHA-256 hash over exactly what was decided and anchored in a block. One call shows whether it has been changed since.

Fraud prevention

Turned away once. Recognized everywhere.

Fraudsters come back with another shipment, another vendor, or inside help. Omnareth remembers who each refusal implicated, opens a case when a pattern appears, and once your analyst confirms it, refuses that identity everywhere.

Detect

Repeat offenders

A driver, vessel, bank account or domain implicated in two denials in 30 days opens a case and goes on watch.

Detect

Schemes across records

One bank account pushed onto three vendors, one release document used twice, one driver working several shipments.

Block

Confirmed fraudsters

Refused everywhere, even after an insider adds them to a dispatch list or marks their account verified.

Fair

Innocent parties

The customer or vendor a fraudster names is never held against anyone. False positives are cleared in one step.

Evidence

A record nobody can quietly rewrite

Each decision stores the record's state, the event, the exact rules evaluated and what was known about everyone involved under an integrity hash, and anchors that hash with the decision in a block.

We tested it the way an insider would attack it, by editing the database directly. Flipping a DENY to ALLOW, editing a seal number, and editing it then recomputing the stored hash to hide the change: all three failed verification.

GET /v1/audits/aud_…/verify
{
  "recordIntact": true,
  "chain": {
    "status": "match",
    "onChainDecision": "DENY",
    "blockHeight": 4
  },
  "verified": true
}

Pilot program

Prove it on your own operation in 90 days

One lane, facility or process, in shadow mode, with results measured against criteria we agree with you. Qualified companies and organizations can apply.