Feature Third-Party Risk
The Keys to a Payment Vendor's Fed Firewall Are Public. Your Bank Still Owns the Risk.
An extortion group leaked firewall keys and credentials for Finzly's Federal Reserve connection. What third-party risk teams must demand now.
Table of Contents
TL;DR
- On October 30, 2026, American Banker reported that an extortion group published passwords, keys, and firewall configuration for the connection that payment vendor Finzly’s bank clients use to reach the Federal Reserve.
- Eleven listed institutions share one connection to ACH, Fedwire, and FedNow. That architecture is explicitly contemplated by the Federal Reserve Banks’ Operating Circular 5, which also says vendor use “shall in no way affect or diminish any obligation or duty of the Institution.” Their failure is your finding.
- The rotation happened. The evidence did not get published. That gap — not the leak — is the thing to close before your next exam cycle.
- The technical root is a five-year-old default: Fortinet’s 2019 advisory on a hard-coded key, with password-based configuration encryption disabled by default (FG-IR-19-007).
Nobody is claiming money moved. That is the least interesting part.
On October 30, 2026, American Banker’s Carter Pape reported that an extortion crew calling itself n0n had published the keys to the firewall guarding Finzly’s link to the Federal Reserve. Finzly is a Charlotte payments company whose BankOS platform connects banks to ACH, Fedwire, and FedNow. The published files are reported to include passwords, keys, and firewall settings, with one link in the configuration labeled “FED” and rules for FedNow and FedACH traffic.
Finzly’s response, as reported: a single device in a backup data center was accessed, system keys and credentials were rotated, there is no evidence client systems or customer data were compromised, and law enforcement was notified. The Fed said payment services were not impacted.
Every one of those statements may be true. Here’s the problem for the eleven institutions listed as Finzly clients: true and evidenced are different products, and only one of them survives an examination.
What actually leaked, and why the shape matters
A stolen password is bounded. You know which account, you disable it, done. A stolen configuration is not bounded, because a firewall config is a map: interface definitions, address objects, policy order, and the encrypted credential strings for everything the device talks to.
That is why the Fortinet advisory matters more than the breach headline. FG-IR-19-007, published November 19, 2019 and revised through February 2024, covers CVE-2019-6693: a hard-coded cryptographic key used to encrypt credential data in FortiOS, FortiManager, and FortiAnalyzer configurations. Fortinet’s own text is unusually direct — “if the CLI configuration is exposed… it is possible to decrypt the encrypted ENC type data to plaintext using this hard-coded cryptographic key.”
The fix requires an administrator to switch on private-data-encryption. Fortinet documents the default in a code comment: /* disabled by default */. The stated workaround is to always password-protect the backup file.
So the exposure chain is mundane. Config backup leaves a device, or an old copy sits somewhere, and unless someone enabled a non-default setting and protected the file, the secrets inside are readable. A technical write-up by CloudSkope reports that the published material was annotated as a 2023 config decrypted October 15, which puts a roughly three-year-old file in play in 2026. Treat that detail as reported rather than confirmed by Finzly.
The backdrop is worse than one vendor. On October 6, 2026 — nine days before that decryption — the FBI and Secret Service issued joint advisory JCSA-20261006-01 on FortiBleed, a credential campaign against internet-facing FortiGate devices, with SOCRadar counting more than 86,644 compromised devices across 194 countries and access flowing to INC/Lynx and Payload ransomware affiliates. The advisory’s mitigations are the boring ones: lock down management access, kill all admin and VPN sessions, reset credentials, require phishing-resistant MFA, and diff every configuration against a known-good baseline.
That last item is the one nobody does. Comparing a running config against a baseline is how you find the account an intruder added and the policy they relaxed. Rotating keys without diffing means you may be rotating a config that already has a hole in it.
The shared connection is the real concentration
American Banker reports that Finzly’s clients reach the networks through a single connection they share. Banks have been outsourcing onto shared pipes for decades and mostly stopped thinking about it, but the Fed’s rulebook never did.
Operating Circular 5 defines “Access Control Feature” to include software, encryption keys, logon IDs, passwords, digital certificates, VPN devices, routers, and tokens. It contemplates the shared arrangement directly: a service provider “may use the same Access Control Features to act on behalf of other Institutions.” It requires prior written notice before an institution or provider shares a connection with another entity. And it closes the accountability question in one line: use of a service provider “shall in no way affect or diminish any obligation or duty of the Institution.”
Read that as a control design instruction. The perimeter you are relying on is not yours, the keys are not yours, and the obligation is.
Three clocks, three different owners
The notification story is where third-party risk and incident response stop being separate functions.
| Clock | Source | Who owes it | Trigger |
|---|---|---|---|
| Immediate notice to the Reserve Bank Support Center, then written confirmation by email | OC 5, §1.4 | Institution or its service provider | Suspected, threatened, or known event that may impact a connection or access control feature. No materiality filter, no hour threshold |
| As soon as possible | 12 CFR 225.303(a) | The bank service provider, to each affected customer | Provider determines an incident that disrupted or is likely to disrupt covered services for four or more hours |
| Within 36 hours | 12 CFR 53.3 | The banking organization, to its federal regulator | Institution determines a notification incident occurred |
OC 5’s language is the strictest of the three and the easiest to trip: “immediately,” triggered by the possibility of impact to hardware that interfaces with the connection. A firewall on the path whose configuration an extortion group says it holds is hard to read as anything else.
The public record does not say when Finzly called the Support Center, or when it told each client. That silence is the exact fact each bank’s own incident file should contain and currently may not. OC 5 also lists the Support Center as 833-FRS-SVCS (833-777-7827) in §1.4 while §5.1(e)(iii) shows 833-377-7827 — put the number in your vendor contact sheet anyway, and confirm the working one before you need it.
Pull these clauses this week
If your contract with any payments vendor looks like the standard paper, it is probably silent on most of this. Five provisions to locate, and what to do if they are missing:
| Clause | What it needs to say | Why now |
|---|---|---|
| Secret and configuration lifecycle | Vendor rotates credentials and configuration secrets on a defined cadence and after any exposure, with written confirmation of completion | ”We rotated” is an assurance. A cadence is a control |
| Evidence delivery | Vendor delivers rotation evidence, config-diff results, and baseline deviations to customers on request, within a stated number of business days | Without this, your file contains a vendor statement and no artifact |
| Shared-connection disclosure | Vendor identifies which customers use the shared Reserve Bank connection and for which payment types | Determines whether you can even scope your own exposure |
| Notification flow | Vendor notifies your named contact and confirms whether it notified the Support Center, with the date and time | Two obligations, one event; you cannot assume theirs was met |
| Subcontractor and fourth-party reach | Vendor names any party with access to its edge configuration, and any managed service provider holding admin rights to it | Config access is the new privileged access |
Two of these are worth escalating rather than negotiating politely. A vendor that will not disclose who shares the connection is asking you to accept unseen concentration, and a vendor that will not commit to evidence delivery is asking you to take its word in front of an examiner. Both belong in the risk acceptance record with a named owner and an expiry date rather than a note in the file.
What proves the rotation
The test for anyone who says the keys were rotated: ask for the artifacts that make rotation verifiable rather than asserted. Each of these either exists or it doesn’t.
- A timestamped rotation log covering every secret class in the exposed configuration — not just the ones named in the news story. VPN pre-shared keys, SNMP strings, API tokens, and certificates negotiated with each client and with the Reserve Bank all count.
- Invalidation proof. The old credential fails and the new one succeeds. A screenshot of a successful new login proves half of it.
- A config diff against a known-good baseline, per device, showing what changed and what was reviewed.
- Session termination logs confirming all admin and VPN sessions were killed, not just re-authenticated. Live sessions survive a password change.
- The IOC sweep, run against your own telemetry using the advisory’s indicators: the C2 and proxy addresses, the beaconing ports, and the eighteen account names in the advisory’s Table 2 —
admin,fortiAdmin,forticloud-sync,Technical_support, and the rest. - The Support Center notification record with date and time, on both sides of the relationship.
If a vendor can produce four of six on request, the program is real. If it can produce the first one and describes the rest, what you have is a press statement with a rotation verb in it.
Where ownership actually sits
This is the part that stalls in the room. Nobody’s job description says “own the vendor’s edge configuration secrets.” At a mid-size bank it usually lands with the TPRM lead by default, who then has no authority over the vendor’s infrastructure team. At a fintech it lands with whoever owns the SOC, who may not be in the vendor conversations at all.
Decide it explicitly and write it down: the TPRM owner holds the evidence file and the escalation path, the security owner holds the technical validation and the diff review, and a named executive holds the risk acceptance if the vendor cannot produce artifacts. Three people, three artifacts, one file. The failure mode is a meeting where everyone assumes the vendor handled it.
30/60/90
Days 1–30 — scope and evidence request
- Identify every vendor with a shared or Fed-adjacent connection. Finzly clients first, then the same architecture at every other processor, core, and payment provider you use.
- Send the evidence request from the list above with a ten-business-day deadline, addressed to the vendor’s risk or compliance contact rather than your account manager.
- Re-tier affected vendors. Shared access control on a Reserve Bank connection should push a vendor to at least High regardless of revenue exposure.
- Open an issue per vendor with a remediation date, so the file shows tracking rather than correspondence.
Days 31–60 — test what you were given
- Run the advisory IOCs against your own logs for June through November 2026, not just the weeks after publication.
- Have security diff the vendor-supplied configuration baseline against what is running now, where the vendor will allow it.
- Validate your own 36-hour path: whoever would draft the notification incident determination should be named in the plan, not discovered during the event.
- Confirm the Support Center number and email ([email protected]) sit in the contact sheet with the OC 5 §1.4 citation beside them.
Days 61–90 — make it repeatable
- Add a configuration-secret rotation clause to the paper at renewal; do not wait for the next full renegotiation.
- Add a standing quarterly question to the vendor review: what changed in your edge configuration and what did you diff it against?
- Tabletop one scenario — attacker holds a config backup and is quiet — and test whether notifications fire on suspicion rather than confirmed loss.
The takeaway
The uncomfortable lesson in the Finzly incident is that the vendor did the expected things. It rotated, notified law enforcement, and told customers. What it did not publish, and what no institution can currently prove for itself, is the dated evidence — which secrets, rotated when, verified how, invalidated against what.
That is a solvable problem, and it is a contract problem before it is a technical one. Institutions that add evidence-delivery language this quarter will get artifacts from every critical vendor at the next event. Institutions that don’t will get a statement and spend a week reconstructing exposure from a news article.
The Third-Party Risk Management (TPRM) Kit has the pieces this work needs: a five-factor tiering rule with a Critical floor for processor and sponsor-bank relationships, a 26-provision contract review against the 2023 Interagency Guidance, and a reassessment scheduler that shortens as residual risk rises.
Related: the September 2026 interagency TPRM proposal, vendor concentration analysis examiners ask for, and forensic evidence preservation in the first four hours.
◆ Need the working template?
Start with the source guide.
These answer-first guides summarize the required fields, evidence, and implementation steps behind the templates practitioners search for.
◆ Related template
Third-Party Risk Management (TPRM) Kit
Tier every vendor with one documented rule, run due diligence and AI vendor reviews, find contract gaps against the 2023 Interagency Guidance, and track SOC reports, issues, concentration and exits. Excel workbook, guide, Word policy and offboarding checklist.
◆ Immaterial Findings · Weekly
Sharp risk & compliance insights. No fluff.
◆ FAQ
Frequently asked questions.
What leaked in the Finzly incident?
Does the bank or the vendor own the notification obligation?
What should a bank ask its payment vendor this week?
Is a rotated key enough?
Does this change how banks should tier payment vendors?
Author
Rebecca Leung
Rebecca Leung has 8+ years of risk and compliance experience across first and second line roles at commercial banks, asset managers, and fintechs. Former management consultant advising financial institutions on risk strategy. Founder of RiskTemplates.
◆ Related framework
Third-Party Risk Management (TPRM) Kit
Tier every vendor with one documented rule, run due diligence and AI vendor reviews, find contract gaps against the 2023 Interagency Guidance, and track SOC reports, issues, concentration and exits. Excel workbook, guide, Word policy and offboarding checklist.
◆ Keep reading
Related posts.
Third-Party Risk
The EBA Just Finalized Its Third-Party Risk Guidelines for Non-ICT Services. If You Thought DORA Covered Your EU TPRM Program, Here's What You Missed.
EBA/GL/2026/09, finalized September 18, 2026, establishes binding third-party risk management requirements for non-ICT arrangements at EU banks and investment firms — the services DORA deliberately excludes. Facilities management, HR outsourcing, legal services, and dozens of other non-technology relationships now fall under a tiered, risk-proportionate framework with two years to comply.
Oct 11, 2026
Third-Party Risk
The Banking Agencies Just Proposed to Throw Out the TPRM Checklist. Here's What the New Risk-Based Framework Actually Expects.
On September 11, 2026, the OCC, Fed, FDIC, and NCUA jointly proposed to replace the 2023 interagency TPRM guidance with a risk-based framework that eliminates mandatory contract terms and reduces requirements for lower-risk relationships. Comment period closes November 16, 2026.
Oct 5, 2026
Third-Party Risk
Federal Regulators Just Listed the Specific Contract Terms in Your Core System Agreement That Will Draw Examiner Scrutiny. Here's What to Check.
On September 11, 2026, the OCC, Federal Reserve, and FDIC issued a joint statement naming four specific contract practices used by core service providers that regulators intend to scrutinize directly. Deconversion fees, back-billing windows, opaque pricing, and integration restrictions are now on the examiner checklist. Here's how to review your core provider agreement before they do.
Oct 4, 2026