A crafted PostScript or PDF document can trigger a privileged write and memory corruption when a vulnerable renderer processes it. The affected upstream Ghostscript releases are those before 10.08.0.Researcher Paulos Yibelo calls the flaw Upload2Shell. His 17 November 2026 report describes controlled command-execution tests through document and image-processing applications. Those tests establish possible exposure paths, not malicious exploitation in the wild.
What it allows
How Does CVE-2026-125396 Escape Ghostscript's Sandbox?
Ghostscript uses the privileged .forceput operation inside a Type 1 font helper. Artifex's patch explains that the helper failed to validate operand types before a copy operation. A crafted document could cause an error, then use its error handler to reach the privileged operation.
The researcher showed that this write could leave a reference to an object after a save and restore sequence reclaimed it. Reusing that memory gave the document control over native interpreter state. In controlled tests, the researcher changed Ghostscript's path-control state and crossed the -dSAFER file and command boundary.
This is a remote code execution risk when an application passes attacker-controlled content to a vulnerable renderer. The CVSS local vector describes the interpreter entry point, not every application path. The researcher demonstrated remote delivery through email previews, uploads, and a shared print queue under specific configurations.
Which Processing Paths Need Attention?
A file extension alone does not establish what the renderer will process. The research traced ImageMagick, Pillow, and other image libraries into Ghostscript during preview or conversion. Merely storing a file, or identifying it without rendering pixels, did not start Ghostscript in those tests.
Roundcube message viewing, WordPress media previews, and CUPS print conversion were separate controlled paths in the report. Each path depended on its application's access settings and installed rendering stack. The command, if reached, ran with the renderer's service account permissions.
The researcher published a self-contained proof of native memory control against Ghostscript 10.07.0. His application demonstrations went further and reported command execution in controlled environments. This distinction matters when evaluating the public proof against an individual deployment.
From 23 August through 21 November 2026, Brandefense Blackhole counted one dark-chat post matching CVE-2026-125396. The measured unit is a post, not an intrusion or an exploited host. This research signal cannot establish active exploitation.
Who is affected
Which Ghostscript Versions Are Affected by CVE-2026-125396?
The VulnCheck CNA identifies upstream ghostpdl releases before 10.08.0 as affected. Artifex's operand-validation patch entered the source before the 10.08.0 release. The researcher observed that his proof failed on the fixed 10.08.0 build.
Distribution packages may carry a backport while retaining an older upstream version number. Conversely, a patched host package does not replace a separate Ghostscript binary in a container or application bundle. Identify the executable or library used by each document worker, then check its vendor package advisory.
An exposed workflow must actually pass untrusted PostScript or PDF content into vulnerable Ghostscript. The researcher found that some wrappers inspect file content and invoke Ghostscript during thumbnails, previews, imports, or print conversion. Other tested decoders did not invoke it for the same inputs.
Use the sandbox definition to distinguish Ghostscript's internal -dSAFER controls from host isolation. Bring the running renderer and its package status into the Vulnerability Intelligence inventory. This inventory step is our defensive inference from the researcher and vendor evidence.
How Should Teams Verify the Fix?
Update upstream Ghostscript to 10.08.0 or a later release, or install a vendor package that explicitly includes the CVE-2026-125396 fix. After deployment, verify the running renderer inside every worker, container, and bundled application. A host package version alone is not sufficient evidence for those copies.
Review which users can submit documents to conversion and print paths. Require authentication on shared print queues where appropriate, and restrict unnecessary upload or preview routes. These measures reduce reachable paths while the renderer is being updated.
Timeline
- 2026-09-07
Artifex committed operand validation for the vulnerable Type 1 font helper.
- 2026-09-08
Artifex released Ghostscript 10.08.0 with the root fix, according to the researcher.
- 2026-11-09
The VulnCheck CNA published CVE-2026-125396.
- 2026-11-17
Paulos Yibelo published the Upload2Shell research and a controlled native-memory proof of concept.
- 2026-11-20
CISA KEV catalog version 2026.11.20 did not list the CVE.
What to do
-
Find the actual renderer
Inventory Ghostscript executables and linked libraries used by mail previews, image conversion, document imports, and print workers. Record the deployed package or bundle for each path.
-
Patch the vulnerable copy
Upgrade upstream Ghostscript to 10.08.0 or later. For distribution builds, use the vendor advisory to confirm a backported fix for CVE-2026-125396.
-
Verify after rollout
Check the version or package release of the renderer each worker actually loads. Confirm that container images and bundled copies were rebuilt or replaced.
-
Reduce reachable paths
Review public upload and preview features and require authentication on shared print queues where appropriate. Keep document workers at the minimum service privileges needed.
-
Investigate local anomalies
Inspect available conversion, mail-preview, and print-job records for unexpected renderer failures or child processes. Treat such findings as leads for investigation, not proof from the public research.
What Brandefense can do
- Watch your stack for exploited flaws
Name the vendors and products you run, and Brandefense alerts you when a CVE in them is exploited.
- See CVE-2026-125396 in Brandefense Atlas
The Atlas record of CVE-2026-125396: severity, exploitation status and the sources behind it.
- Vulnerability intelligence platform
Brandefense Vulnerability Intelligence follows which vulnerabilities attackers are using, so your team can patch what is under attack first.
Frequently Asked Questions
Does CVE-2026-125396 require local access?
The CNA rates the Ghostscript flaw with a local attack vector. The researcher demonstrated remote delivery when an application accepted a document and passed it to vulnerable Ghostscript.
Is Ghostscript 10.08.0 the fixed upstream release?
Yes. The CNA marks earlier upstream releases affected, and the researcher says Artifex released the root fix in 10.08.0. Distribution backports need package-specific verification.
Does updating a host package fix every application?
No. A container image or application bundle may still carry its own Ghostscript copy. Verify the renderer that each file-processing worker actually uses.
Has malicious exploitation been confirmed?
The cited research reports controlled proof-of-concept tests. The public evidence used for this profile does not establish malicious exploitation at a specific deployment.
Sources
- How We Hacked the Pipes: Upload2ShellPaulos Yibelo, pwn.ai
- CVE-2026-125396 RecordVulnCheck CNA via CVE Program
- Known Exploited Vulnerabilities CatalogCISA