Ghostscript sandbox escape illustrated by a document breaching a renderer’s inner security barrier

Ghostscript sandbox escape: Find and secure exposed apps

Key takeaways

  • CVE-2026-125396 lets a crafted document bypass Ghostscript’s sandbox and execute commands as the rendering account.
  • Researchers reached the vulnerable renderer through upload forms, email processing, and print queues, including some unauthenticated paths.
  • Ghostscript 10.08.0 contains the upstream fix, but administrators must verify the renderer inside each application, worker, and container.
  • Application patches, restricted input formats, authentication, and operating-system confinement provide separate layers of protection.

A Ghostscript sandbox escape, tracked as CVE-2026-125396, can turn a crafted uploaded file into command execution on the server that renders it. According to The Hacker News, Ghostscript 10.08.0 fixes the flaw, but administrators must update the actual renderer behind each application—not just the host operating system.

In its November 22, 2026 report, The Hacker News described pwn.ai’s demonstrations across about ten applications. Ordinary preview, upload, and print features became delivery routes to a vulnerable interpreter, sometimes without an attacker logging in.

The report said there were no reports of exploitation in attacks. The practical priority is therefore exposure discovery and verified remediation: identify which workflows reach Ghostscript, establish their patch status, and contain rendering workers while fixes reach every copy.

How does the Ghostscript sandbox escape work?

Ghostscript interprets documents for rendering and conversion. According to The Hacker News, many Linux servers use it to process PDFs, PostScript files, and images, including through applications that never invoke it directly.

The flaw is a use-after-free in font-loading code. In plain language, the interpreter mishandles memory associated with an object it has already released. A crafted document exploits that mistake to disable checks enforced by -dSAFER, Ghostscript’s sandbox intended to prevent documents from running commands or accessing files.

In pwn.ai’s demonstrations, the result was server-side command execution under the low-privilege account performing the rendering. That is not the same as demonstrated administrator or root access. Nevertheless, the security boundary between an untrusted document and server execution had failed.

The important distinction is between the vulnerable component and its delivery route. Although the CVE is classified as local, researchers reached it over the network through applications accepting files, email attachments, and print jobs. A local classification should not substitute for examining how your application invokes the component.

Which applications and upload workflows need review?

The Hacker News identifies ImageMagick, Pillow, libvips, and a Qt plugin as possible intermediaries that pass PostScript or EPS content to Ghostscript. The exact route depends on the libraries and configuration used by the application.

A file extension is not a dependable boundary here. The report says this handoff can occur even when the file is named .png, .jpg, or .pdf. A feature advertised as image cropping or thumbnail generation may therefore deserve the same scrutiny as an explicit document-conversion service.

The reported examples illustrate different access requirements:

Reported workflow Access needed in the demonstration Review priority
Shared Debian 13 print queue Anonymous print submission Check queue authentication and renderer patch status
Roundcube attachment processing An emailed attachment, processed when the recipient opened the message Review attachment rendering and worker isolation
Forminator and Essential Addons for Elementor Public upload forms Trace upload processing to downstream libraries
WordPress Core An account permitted to upload files; Authors and higher by default Verify both WordPress and Ghostscript fixes

These are demonstrated routes, not proof that every installation of each product is exploitable. Configuration, rendering dependencies, and patch status still need checking locally.

Recommended assessment: prioritize workflows that accept untrusted files and automatically generate previews, thumbnails, converted documents, or print output. Do not limit the inventory to endpoints whose names mention PDF or PostScript.

Our guide to hidden file metadata and its security implications offers useful context for reviewing more than a file’s visible appearance. For this vulnerability, however, removing metadata is not a substitute for fixing the renderer.

How to identify the renderer your application actually uses

The most consequential operational warning in the report is that updating the host does not update a separate Ghostscript copy bundled into an application or container image. A successful package update is therefore evidence about one installation—not your whole processing chain.

Use the following as an assessment checklist rather than an exploitation test:

  1. Map entry points. List upload forms, attachment previews, image-editing functions, conversion jobs, and print queues. Record whether each requires authentication.
  2. Trace processing. Review application configuration, dependency information, and worker definitions to determine whether those functions can invoke Ghostscript indirectly.
  3. Locate each runtime copy. Inspect the environments that perform the work, including containers and separate document workers. Record the executable location, package source, and reported version.
  4. Establish patch evidence. Compare the installed build with the upstream fix or the package vendor’s security information. Where the package status is unclear, treat the result as unresolved rather than assuming safety.
  5. Record the execution boundary. Identify the rendering account, accessible directories, credentials available to the worker, and any operating-system confinement.

Keep the resulting inventory concrete: application, feature, worker environment, Ghostscript build, patch evidence, and owner. That turns a broad vulnerability alert into an actionable remediation list.

For administrators tracing executables and application directories, this Linux file-system guide provides background. The official Ghostscript website is a useful starting point for upstream information, while the ImageMagick website supports dependency review where that processor is involved.

Which patches address the vulnerability?

According to The Hacker News, Artifex fixed CVE-2026-125396 in Ghostscript 10.08.0 on September 8, 2026. The report gives the vulnerability an advisory score of 8.5 out of 10.

WordPress addressed its own upload route in version 7.0.4 on August 12, 2026, under the separate identifier CVE-2026-65640. These fixes address different points in the chain. Updating WordPress does not remove the need to fix Ghostscript, and updating Ghostscript does not replace application maintenance.

Distribution coverage also requires care. The survey cited in the November report listed Debian 12 as vulnerable; Amazon Linux 2, 2023, and 2027 Preview as awaiting fixes; and several SUSE and openSUSE releases as affected or undergoing remediation. Those were package-status observations at the time, not tested exploitation paths for every distribution.

The research did not establish fix status for current Red Hat, Rocky Linux, AlmaLinux, Oracle Linux, or stable Fedora packages. An omitted or unconfirmed distribution should not be interpreted as unaffected.

Recommended remediation: obtain the appropriate fixed build for each environment, update bundled copies as well as system packages, redeploy affected workers, and verify the runtime afterward. Preserve the evidence showing which build now processes files. Treat historical distribution lists as investigation leads, not current patch instructions.

What protections help while you patch?

The report identifies three useful measures: requiring authentication on print queues, restricting formats accepted by image processors, and applying operating-system controls. In one demonstration, an AppArmor profile prevented the compromised interpreter from starting a shell.

That is evidence of a blocked action in one test, not proof that every AppArmor profile prevents every consequence. Likewise, authentication reduces anonymous reachability but does not repair vulnerable rendering code.

For a layered interim response, consider these recommendations:

  • Restrict unnecessary processing. Disable unused document-conversion and preview features where the operational impact is acceptable.
  • Enforce format policy at processing time. Check what the rendering stack accepts rather than relying only on filenames or browser-supplied labels.
  • Reduce worker privileges. Review whether rendering accounts need access to application secrets, sensitive directories, or unrelated services.
  • Apply confinement. Review operating-system restrictions around worker processes and test that legitimate rendering still works.
  • Limit anonymous entry points. Require authentication for shared print queues and reconsider public conversion features until remediation is verified.

These are defense-in-depth recommendations, not alternative fixes. For broader security-program guidance, CISA provides an authoritative reference point.

Be equally deliberate about validation. The Hacker News says one published proof-of-concept demonstrates the core memory manipulation without running shell commands or touching host files, and fails before that point on patched 10.08.0. That narrower demonstration should not be confused with the broader application-level command-execution tests.

Avoid submitting exploit samples to production services. If your security team needs behavioral testing, use an authorized, isolated environment with a defined scope; our introduction to safe malware analysis explains the general safety mindset.

The Bottom Line

The Ghostscript sandbox escape is a reminder to patch the component that actually handles untrusted content. A protected front end or updated host cannot establish that a separate rendering worker is fixed.

Start with your most exposed upload, attachment-preview, or print workflow. Trace it to the running renderer, verify the applicable Ghostscript fix, and document any remaining application updates or containment gaps. The next step is not simply closing a patch ticket—it is proving that the vulnerable processing path has been removed.

Frequently Asked Questions

What is the Ghostscript sandbox escape vulnerability?

CVE-2026-125396 is a use-after-free flaw in Ghostscript’s font-loading code. According to The Hacker News, a crafted document can use it to disable SAFER sandbox checks and execute commands under the rendering account.

Which Ghostscript version fixes CVE-2026-125396?

Artifex fixed the vulnerability in Ghostscript 10.08.0 on September 8, 2026, according to The Hacker News. Administrators should verify the actual build used by every application or worker, including bundled and containerized copies.

Can an image upload trigger the Ghostscript vulnerability?

Yes, in processing configurations that pass the uploaded content to Ghostscript. The reported research showed that intermediary libraries could hand PostScript or EPS content to the interpreter even when the filename ended in .png, .jpg, or .pdf.

Does updating WordPress fix the Ghostscript flaw?

WordPress 7.0.4 closed its own upload path under CVE-2026-65640, according to the report. That application fix and the Ghostscript fix address separate points in the processing chain, so administrators need to verify both.

Was CVE-2026-125396 exploited in real attacks?

The Hacker News reported no known attacks in its November 22, 2026 coverage. Researchers nevertheless demonstrated command execution through multiple application workflows, including some that did not require authentication.

Is updating the host operating system enough?

Not necessarily: an application or container can contain a separate Ghostscript installation. Check the renderer inside each processing environment and verify that the fixed build is the one actually running.

Sources

Enjoyed this? Subscribe to the InnoVirtuoso newsletter for practical AI, security and tech insights — and share your thoughts in the comments.

Browse InnoVirtuoso for more!