Skip to main content

Test the beneficiary record before trusting a payment approval

A staging exercise for beneficiary changes, approval binding and independent reconciliation after CERT-In's 20 November advisory.

Published
Reading time
2 min
On this page (4)

A payment can be correctly approved against a record that was changed without permission. That is the testing problem raised by CERT-In's 20 November advisory CIAD-2026-0062. Its observations extend beyond financial institutions to service providers supporting government-linked payments and vendor or partner payouts.

The following is a proposed assessment exercise, not a reconstruction of any disclosed victim's system.

Start with the beneficiary record

Use a staging workflow, synthetic invoices and accounts controlled by the test team. Agree the scope and stop conditions with the application owner. No real payment should leave the environment.

Create a legitimate supplier and an invoice. Capture the supplier identifier, destination account, amount and record version. Ask the business owner to identify the evidence a real approver sees. A supplier name alone tells the reviewer very little about where the money will arrive.

Next, give the tester the access available to a compromised provider administrator. Attempt to change the destination account without changing the supplier name. Follow the application through approval and simulated execution. Does the approver see the changed destination? Does any independent check compare it with a trusted supplier record?

A rejection is useful only if the team can show which boundary rejected the change. A failed screen interaction does not establish that another administrative path cannot make it.

Change the record after approval

Repeat the exercise with a different sequence. Obtain approval for the original destination, then attempt to change the record before the queued transaction executes.

Inspect what approval binds to. An approval attached only to an invoice number may remain valid while the destination behind it changes. An approval tied to the destination, amount and record version can instead become invalid when those inputs change. Test the actual implementation; do not assume either design.

Retain the approved view, the record-change log and the simulated payment instruction. The comparison should let an engineer explain exactly what was approved and what would have executed.

Test the reconciliation source

An internal comparison can look reassuring when both sides read the same altered database. Map who can write each record used for reconciliation, including administrators and service accounts.

Try the comparison again using a separately controlled reference. The test should establish whether the compromised provider role can change that reference too. A second dashboard is not independent merely because a different team opens it.

For a multi-client provider, include a scoped tenant-boundary test. Check whether one client's support or administrative access can reach another client's payment records. Use separate synthetic tenants and capture denied as well as successful operations.

Hand over a result the owner can reproduce

The assessment deliverable should include the starting permissions, record versions, approval evidence, attempted mutations, simulated execution outcome and the control that accepted or rejected each step. Keep sensitive account data out of screenshots and reusable training material.

A useful retest repeats the failed sequence with the same permissions. It demonstrates that changing a destination invalidates approval, or that an independent verification step blocks execution. A screenshot showing another approval box does not establish that outcome.

The advisory also discusses automated intrusion activity separately. It does not establish that those tools caused the payment cases. It introduces no new fixed deadline for commissioning this assessment. Its incident-reporting reminder refers to existing CERT-In Directions.