The Patch Doesn’t Delete the Log: What a Badge System Installer Left Behind

Amigos, this one isn’t a flashy remote exploit. Nobody’s popping a shell from the internet. And honestly, I think that’s why it’s easy to get wrong.

On November 5, Johnson Controls published JCI-PSA-2026-45 for CVE-2026-106019. The short version: certain installers for C-CURE 9000 and victor, the physical security platforms a lot of plants, campuses and government sites use for doors, badges and cameras, could write administrative credentials into plain text log files during install.

The fixed installer is out. Great. But mira, here’s the part I want you to sit with: a fixed installer doesn’t reach back in time and erase the logs the bad one already wrote. It doesn’t change the passwords that are sitting in them either.

What actually happened

Per Johnson Controls’ advisory, the C-CURE 9000 3.20 installer at build 3.20.597 and the victor 8.0 installer at build 8.0.597 “may write sensitive authentication information in cleartext” into local installation log files. That can happen during an install, an upgrade or a repair.

JCI describes what gets logged as “administrative credentials or other sensitive authentication data.” The NVD record, which came from JCI, frames it as allowing “Use of Known Operating System Credentials.” I’m going to stick to that language. The advisory doesn’t spell out every field in every log, so I won’t guess.

The affected builds were available for download from July 1 through September 18, 2026. The fixed builds are 3.20.607 or later for C-CURE 9000, and 8.0.607 or later for victor.

The logs live under C:\ProgramData\JCI\InstallerTemp\. JCI lists nine specific files, including UnifiedInstallDashboard.log, UnifiedServer_[TIMESTAMP].log and VictorApplicationServer_[TIMESTAMP].log. If you run these products, go read the full list in the advisory. I won’t explain how to dig passwords out of them.

The bug class is CWE-532, Insertion of Sensitive Information into Log File. JCI scores it 8.8 High on CVSS 4.0. There’s no CVSS v3.1 score published, so if you see one floating around, it didn’t come from JCI or NVD.

“It’s just a local log” isn’t the full story

It’s tempting to shrug this off as a local-only issue. Two problems with that.

First, the scoring itself. JCI’s vector is AV:A, adjacent network, with low privileges required and attack requirements present. So it isn’t scored as local-only.

Second, and more important for real life, JCI’s own advisory lists harvesting from log aggregation, monitoring and backups as a way in. Think about where install logs actually go in a mature shop:

  • Your SIEM collector picks them up.
  • Your backup job snapshots the server, logs and all.
  • Someone zips up a support bundle and sends it to the integrator.
  • A VM snapshot sits on a storage array for a year.

Every one of those is a copy. So the practical exposure is anyone who can read those logs anywhere they’ve traveled, not just someone sitting at the console.

It’s like writing the alarm code on a sticky note, then photocopying the sticky note for the file cabinet, the offsite archive and the contractor’s binder. Throwing away the original note doesn’t help much.

Why this matters for OT

Let me be clear about what this is. C-CURE 9000 and victor are physical access control and video management. They’re not SCADA, not an EMS, and this isn’t a grid exploit. It’s credential exposure, not remote code execution.

But claro, in a plant, a utility, an airport or a data center, the badge system decides who can walk into the control room, the substation yard, the wiring closet and the engineering workstation room. Physical access control is part of the OT attack surface, full stop. If someone can get into the room with the engineering laptop, a lot of your network segmentation stops mattering.

And these systems rarely live alone. They often share a campus, identity integrations and sometimes network trust with the building management system and OT support networks. So ask yourself the uncomfortable question: was the admin account typed into that installer used anywhere else? Is it a domain account? Is it federated? If the answer is “maybe,” treat it like it’s out there.

We’ve seen this movie before

Here’s what bugs me. This isn’t new for this product family.

Back in May 2020, CISA’s ICSA-20-142-01 covered CVE-2020-9045: C-CURE and victor installs and upgrades storing sensitive information in cleartext, with a persistent install log. The fix guidance back then was to delete the logs from an installer temp folder and change the Windows account password. Sound familiar?

Then in May 2024, ICSA-24-135-03 covered CVE-2024-0912, the same CWE-532 class, where IIS logs on the C-CURE 9000 web server captured Windows credential details.

So, familia, I’m not treating this as a one-off curiosity. Installer hygiene and log retention need to be standing controls for your physical security stack, not something you think about only when an advisory drops.

Where it stands right now

As of November 6, Johnson Controls hasn’t reported any exploitation, and the CVE isn’t in CISA’s Known Exploited Vulnerabilities catalog. I also couldn’t find a CISA ICS advisory for this specific CVE yet, so don’t go looking for an ICSA number that doesn’t exist.

That’s your window. Use it calmly, but use it.

What to do

JCI’s remediation boils down to four things: use the fixed installer, delete local copies of the affected installer, delete the listed log files and rotate any exposed credentials. Here’s how I’d put that to work:

  • Figure out if you’re in scope. Find any C-CURE 9000 3.20 or victor 8.0 host that was installed, upgraded or repaired with build 3.20.597 or 8.0.597, especially between July 1 and September 18.
  • Get rid of the bad installer. Check software distribution packages, integrator shares and technician laptops for the .597 builds and replace them with .607 or later.
  • Delete the logs everywhere they went. Clean the nine files on the host, then search your SIEM, backup catalogs and support bundles for those filenames and the InstallerTemp path.
  • Rotate the credentials. Change the administrative and operating system accounts that could have gone through the installer, including service accounts. Patching alone does not invalidate a password that’s already been written down.
  • Look back. JCI suggests reviewing C-CURE 9000 and victor activity before November 3, 2026, and checking file access audits for non-admin reads of installer temp directories.
  • Harden the folder going forward. Tighten permissions on C:\ProgramData\JCI\, stop shipping raw install logs in support zips, and treat the physical security admin network like an OT conduit with segmentation, MFA and jump hosts.

JCI also reminds customers to test changes outside production first. In OT that’s not optional, por favor.

Bottom line

  • The patch fixes the installer, not the past. Old logs and exposed passwords stay until you delete and rotate them.
  • Logs travel. Your SIEM, backups and support bundles may be holding copies of credentials you thought were local.
  • The badge system is part of the OT attack surface. Its admin credentials deserve the same care as your engineering workstation’s.

Stay sharp out there.

Javier

Sources

  1. https://tyco.widen.net/s/jmcjbkpf65/jci-psa-2026-45
  2. https://www.johnsoncontrols.com/trust-center/cybersecurity/security-advisories
  3. https://nvd.nist.gov/vuln/detail/CVE-2026-106019
  4. https://www.cisa.gov/sites/default/files/feeds/known_exploited_vulnerabilities.json
  5. https://cwe.mitre.org/data/definitions/532.html
  6. https://www.cisa.gov/news-events/ics-advisories/icsa-20-142-01
  7. https://www.cisa.gov/news-events/ics-advisories/icsa-24-135-03
  8. https://www.johnsoncontrols.com/trust-center/cybersecurity/resources

Leave a Reply