Ghostscript CVE-2026-125396 SAFER Bypass Lets Crafted Documents Run Code: Update to 10.08.0 Now

Rescana advisory card: Ghostscript CVE-2026-125396 SAFER bypass, CNA CVSS 4.0 8.5 High, no reported exploitation, update to 10.08.0.

Status: Artifex Ghostscript versions before 10.08.0 have CVE-2026-125396, a bypass of the interpreter's -dSAFER restrictions. VulnCheck, the CVE Numbering Authority (CNA), says a crafted PostScript or PDF document can cause a use-after-free and "execute arbitrary code." The CNA scores it CVSS 4.0 8.5 (High) and CVSS 3.1 7.8 (High), both with a local attack vector. The fix is Ghostscript 10.08.0. On November 21, 2026, a post on the oss-security list called attention to this CVE, using the name "upload2shell," and relayed that a research publication that day identified CVE-2026-125396 as also fixed in 10.08.0, consistent with the CNA's affected range. As of 06:49 EST on November 22, 2026, no source we opened reports exploitation in the wild. CVE-2026-125396 is not in CISA's Known Exploited Vulnerabilities (KEV) catalog. The catalog version we checked is 2026.11.20, released at 12:48 EST on November 20, 2026, with 1,790 entries. Debian has fixed packages. Ubuntu still lists every release as "Needs evaluation." SUSE rates the issue important and shows an overall state of Pending. We checked the sources between 06:27 and 06:32 EST on November 22, 2026, and rechecked them between 06:48 and 06:49 EST the same day.

Executive Summary

Ghostscript is the interpreter many servers and desktops use to render, convert or preview PostScript and PDF files. The -dSAFER option is meant to keep document code from writing files or running commands. CVE-2026-125396, as described by the CNA, bypasses that restriction in a font-loading helper, so a crafted document can run code as the Ghostscript process.

What matters for planning:

  • The CNA says a crafted document is enough. VulnCheck's description is that attackers can "supply crafted PostScript or PDF documents" that lead to arbitrary code execution. The CVSS 4.0 vector is local, with passive user interaction: the document has to be handed to Ghostscript. In Rescana's reading, that includes a person opening a file and also any service that renders or converts a file automatically.
  • The 10.08.0 fix is already a few months old. Oss-security and Ghostscript's current release notes date 10.08.0 to September 8, 2026. Ghostscript 10.08.0 or later, or a distribution package with a backport, already contains the fix. Older packages may still be in use.
  • No exploitation in the wild is reported. CISA's Coordinator enrichment in NVD lists exploitation as "none" (recorded 11:29 EST on November 10). CISA has not added the CVE to KEV. We found no public proof of concept in the sources we opened, and this post links to none.
  • Other programs that pass files to Ghostscript are in scope. The oss-security announcement said a research publication the same day "detail[ed] ways it was found to be exploitable through other programs passing files to Ghostscript." We do not reproduce those paths. Treat any service that sends untrusted documents to Ghostscript (preview, conversion, print, mail and upload pipelines) as needing an inventory of its Ghostscript version (Rescana guidance built on that oss-security statement).
  • Distribution coverage is uneven. Debian has fixed packages for bookworm, trixie, forky and sid. Ubuntu's tracker still says "Needs evaluation" on every listed release, and we found no Ubuntu security notice. SUSE's overall state is Pending; openSUSE Tumbleweed already lists a 10.08.0 package, and several SUSE Linux Enterprise lines show packages in QA. Red Hat has a public bug (2569593, New, severity high) and no fixed-in version. The Customer Portal CVE page and data API still returned HTTP 404 when we rechecked them at 06:49 EST.

If you run Ghostscript yourself, or you run software that calls it, find the version, update to 10.08.0 or your vendor's fixed package, and restart anything that loaded the old binary (Rescana guidance). This is a new post: an older Rescana article on other Ghostscript findings from February 2026 does not cover this CVE and is not a follow-up.

Technical Information

The following is the CNA's published description and the vendor's published fix. This post gives no payload details, no reproduction steps and no links to exploit code.

VulnCheck's CVE record, published as CNA, is titled "Ghostscript before 10.08.0 -dSAFER Bypass via .putgstringcopy .forceput Recovery." The description says: "Ghostscript before 10.08.0 contains a -dSAFER bypass vulnerability in the .putgstringcopy procedure in gs_fonts.ps, which calls the privileged .forceput operator without checking operand types. Attackers can supply crafted PostScript or PDF documents that trigger a type error, recover .forceput via a custom error handler, and cause a use-after-free across save/restore to execute arbitrary code." The record lists CWE-648, "Incorrect Use of Privileged APIs." The CNA credits Paulos Yibelo and pwn.ai as the finder, and Ken Sharp as the remediation developer. This post does not link a research site for that credit.

The CVSS 4.0 vector in the record is CVSS:4.0/AV:L/AC:L/AT:N/PR:N/UI:P/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N, base score 8.5, severity High: local, low complexity, no attack requirements, no privileges, passive user interaction, and high impact on confidentiality, integrity and availability of the vulnerable system. The same record also has CVSS 3.1 CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H, base score 7.8, severity High. NVD carries both VulnCheck scores as Secondary and shows vulnerability status Awaiting Analysis when we checked. NVD has not produced a score of its own.

CISA's Coordinator SSVC enrichment in the NVD record lists exploitation "none," automatable "no," and technical impact "total," with timestamp 11:29 EST on November 10, 2026. We do not treat that November 10 field as a statement about later attention to the CVE.

The GitHub commit that Debian's tracker and the CVE record identify as the fix (in Resource/Init/gs_fonts.ps) is titled "PS interpreter - validate operands to .putgstringcopy." The commit message says the change "prevents the problem by ensuring that the operands are the correct types," so that a type error in copy does not give a custom error handler a path to .forceput. The commit's author date is 06:04 EST on September 2, 2026, and the committer date is 06:42 EST on September 7, 2026. Git timestamps are set by the committer and are not proof of when a change became public.

Severity, from whom. The Rescana severity field for this post follows the CNA's High rating (CVSS 4.0 8.5) and attributes it. SUSE also rates the issue "important" and publishes the same CVSS 3.1 7.8 and CVSS 4.0 8.5 figures, labelled as both CNA (VulnCheck) and SUSE. Ubuntu's tracker lists Ubuntu Medium priority, while showing the same CNA scores (High). These are different views. We do not resolve the difference.

Affected Product Versions

Product or component Affected Fixed or status Source
Artifex GhostPDL / Ghostscript Versions from 0 before 10.08.0 10.08.0 CVE record (CNA VulnCheck); VulnCheck advisory
Ghostscript 10.08.0 Not affected (the fixed release) Dated September 8, 2026, by oss-security, by Ghostscript's current News page, and by the git tag commit oss-security; Ghostscript News (latest); git tag
Debian bookworm 10.0.0~dfsg-11+deb12u8 vulnerable bookworm 10.0.0~dfsg-11+deb12u9 (DLA-4914-1); trixie fixed from 10.05.1~dfsg-1+deb13u2 (DSA-6516-1), current security package 10.05.1~dfsg-1+deb13u3; forky and sid 10.08.0~dfsg-2 Debian security tracker
Ubuntu (26.10, 26.04 LTS, 24.04 LTS, 22.04 LTS, 20.04 LTS, 18.04 LTS, 16.04 LTS) Needs evaluation No fixed version and no Ubuntu security notice found Ubuntu CVE page (publication date 10 November 2026, last updated 18 November 2026)
SUSE and openSUSE Overall state Pending; many current and LTSS products Affected or In progress openSUSE Tumbleweed lists ghostscript 10.08.0-5.1 as a released package (patch openSUSE-Tumbleweed-2026-12528). Several SUSE Linux Enterprise 15 and 12 lines show backported packages (for example ghostscript 9.52-150000.222.1) in QA. No SUSE security announcement is cross-referenced. SUSE Bugzilla 1289822 (In progress) SUSE CVE page
Red Hat No Customer Portal CVE page retrieved (HTTP 404 on the first check and again at 06:49 EST) Bugzilla 2569593, product Security Response, status New, severity high, no fixed-in version listed Red Hat Bugzilla (created 08:09 EST, November 10, 2026); Customer Portal data API and HTML page still returned 404 at 06:49 EST
WordPress (separate CVE, see the note below) WordPress 4.7 through 7.0.3 in the branches listed by the advisory, when Imagick and Ghostscript are in use 7.0.4, and backports including 6.9.7, 6.8.8, 6.7.7 and earlier branches back to 4.7.35 WordPress advisory GHSA-8vr3-7mxf-gx8w (CVE-2026-65640), published August 12, 2026; pointed to from oss-security

Two points on this table. First, a distribution's fixed package version may be a backport (Debian's trixie and bookworm packages, and the SUSE packages in QA), not 10.08.0 itself. Second, WordPress advisory GHSA-8vr3-7mxf-gx8w is a separate CVE (CVE-2026-65640). The advisory says certain WordPress installations are vulnerable to remote code execution "via malicious Postscript file upload by an Author level user or higher," with prerequisites that Imagick and Ghostscript are in use, "the weakness is in Ghostscript's handling of certain embedded files," and the uploader has the upload_files capability. A follow-up on oss-security relayed that the same research website says WordPress fixed this issue on August 12. The WordPress advisory itself does not name CVE-2026-125396, so we report that link as relayed on oss-security and do not treat the two CVE IDs as the same record.

The same 10.08.0 release also fixed five other CVEs listed on Ghostscript's CVE index: CVE-2026-110133, CVE-2026-112922, CVE-2026-103226, CVE-2026-19547 and CVE-2026-39919. We retrieved that index at 06:49 EST on November 22, 2026. It lists those five IDs for 10.08.0 and does not list CVE-2026-125396. The oss-security announcement had cited the same five IDs from that index. This post is about CVE-2026-125396. Debian's DLA-4914-1 and DSA-6516-1 also list some of those other IDs among their references.

Timeline (EST)

Source timestamps use UTC, Pacific Time (UTC-8) or Eastern Time (UTC-5). Everything below is converted to EST.

  • March 16, 2026. Ghostscript 10.07.0 is dated in its News page. The git tag commit is at 05:20 EST (git timestamps are set by the committer).
  • August 12, 2026. WordPress publishes advisory GHSA-8vr3-7mxf-gx8w (CVE-2026-65640) and the 7.0.4 / backport fixes (date only, from the advisory page).
  • September 2, 2026, 06:04 EST, and September 7, 2026, 06:42 EST. Author and committer dates on the Ghostscript fix commit for .putgstringcopy (git timestamps).
  • September 8, 2026. Ghostscript 10.08.0 is released, per oss-security, Ghostscript's current News page, and the git tag commit (03:32 EST). The versioned News page at the 10.08.0 docs URL still shows the date 2026-05-19; the 10.07.0 News page is dated 2026-03-16, so that May date is not the 10.07.0 date. Oss-security said the 10.08.0 notes "accidentally list the 10.07.0 release date instead." We report the three September 8 sources above as the release date, and we note the stale date on the versioned docs page.
  • November 9, 2026, 17:03 EST. CVE-2026-125396 is reserved (CVE record).
  • November 9, 2026, 18:28 EST. The CVE record is published by VulnCheck. ENISA's EUVD lists the same publication time as EUVD-2026-120559.
  • November 9, 2026, 19:17 EST. NVD publishes the record (status later shown as Awaiting Analysis).
  • November 9, 2026, 19:31 EST. The GitHub Advisory Database publishes GHSA-hg3g-c775-3fpj as an unreviewed entry.
  • November 10, 2026, 08:09 EST. Red Hat Bugzilla 2569593 is created.
  • November 10, 2026, 11:29 EST. CISA's Coordinator SSVC enrichment is recorded in NVD (exploitation none).
  • November 10, 2026. Ubuntu lists the CVE at Medium priority, with a publication date of 10 November 2026. Ubuntu's page later shows a last-updated date of 18 November 2026, still "Needs evaluation."
  • November 21, 2026, 14:09 EST. Alan Coopersmith posts to oss-security that Ghostscript's CVE index lists five CVEs as fixed in 10.08.0, and that a research publication that day identified CVE-2026-125396 as also fixed in that release. The subject line uses the name "upload2shell."
  • November 21, 2026, 14:59 EST. Jan Schaumann replies on oss-security, pointing to the August 12 WordPress advisory and arguing that a fix dated September 8 means the vulnerability should be treated as public from that date. (The oss-security archive files this message under November 22.)
  • November 21, 2026, 23:06 EST, through November 22, 2026, 01:30 EST. The CVE record, NVD and the GitHub entry are updated (reference and last-modified times; NVD status remains Awaiting Analysis).
  • November 22, 2026, 06:27 to 06:32 EST, and 06:48 to 06:49 EST. We checked the sources listed below, then rechecked them. State at the recheck: not in KEV, NVD Awaiting Analysis, no in-the-wild exploitation reported, Debian fixed, Ubuntu Needs evaluation, SUSE Pending with a Tumbleweed package and SLE packages in QA. The official Ghostscript CVE index, which had not been retrievable at the first check, loaded and listed the five other 10.08.0 CVEs above, not CVE-2026-125396. The Red Hat Customer Portal CVE page still returned HTTP 404. Ghostscript bug 709684 still returned a bot-challenge page and is not cited.

All dates and times in this post are EST.

How to Find Your Exposure

Evidence first. No source we opened publishes indicators of compromise for CVE-2026-125396, so exposure is about inventory (Rescana guidance, except where a vendor version is cited).

  1. Hosts and containers that run Ghostscript. Check the installed gs or ghostscript version (gs --version or the package manager). Anything older than 10.08.0 is affected unless a vendor backport is installed. Rebuild container images from patched base layers, and restart running processes, which keep the old binary loaded. Applications, containers and appliances that ship their own copy of Ghostscript (a bundled or vendored gs) need their own update; a package-manager check on the host will not find those copies (Rescana guidance).
  2. Distribution packages. On Debian, compare with the fixed versions in the table. On Ubuntu, the tracker still said "Needs evaluation" when we checked, so confirm against your Ubuntu security updates rather than assuming a package exists. On SUSE, check whether Tumbleweed 10.08.0-5.1 or a later QA update has been released for your product. On Red Hat, watch bug 2569593 and the security errata feed; we found no fixed-in version.
  3. Applications that call Ghostscript. Many preview, conversion, print and mail tools invoke gs or link GhostPDL to handle PostScript, EPS, PDF or some image formats. Ask each vendor which Ghostscript version their product embeds or calls, including language bindings and ImageMagick or GraphicsMagick delegates (Rescana guidance, extending the oss-security remark about "other programs passing files to Ghostscript").
  4. WordPress sites. If a site runs Imagick with Ghostscript, the WordPress advisory applies in addition to the interpreter update. Check the WordPress version against the patched versions in GHSA-8vr3-7mxf-gx8w, and still update Ghostscript itself (Rescana reading: a WordPress-side check does not replace the interpreter fix).
  5. Where untrusted documents arrive. Make a list of services that convert, preview, print or otherwise pass uploaded or emailed files to Ghostscript. Those are the places a crafted document can reach the interpreter. This is Rescana's prioritization, not a vendor instruction.

Workaround and Mitigation

The published remediation is to run a fixed Ghostscript. VulnCheck's affected range ends at 10.08.0. The oss-security announcement relays a research publication identifying 10.08.0 as the release that includes the fix for this CVE. The fix commit message describes the change as validating operand types in .putgstringcopy.

What that means in practice (Rescana guidance, except where a vendor package is cited):

  • Upstream and Windows builds: install Ghostscript 10.08.0 or later from Artifex's published packages, and confirm the version afterward.
  • Debian: apply DLA-4914-1 on bookworm and DSA-6516-1 on trixie (current security package on trixie is 10.05.1~dfsg-1+deb13u3). Sid and forky already show 10.08.0~dfsg-2.
  • SUSE: Tumbleweed already lists 10.08.0-5.1. For SUSE Linux Enterprise, watch for the packages now shown in QA to be released, and install the vendor update when it is announced.
  • Ubuntu and Red Hat: no fixed package was published when we checked. Watch each vendor's tracker for CVE-2026-125396 and install the update when it appears.
  • After any update: restart every service and application that loaded Ghostscript, including long-running converters, print queues and web applications.
  • WordPress: if Imagick and Ghostscript are in use, apply the patched WordPress versions listed in GHSA-8vr3-7mxf-gx8w, and update Ghostscript as well.
  • If you cannot update soon: stop passing untrusted documents to Ghostscript (disable preview, conversion or print paths that call gs) until the fix is in place. Two further interim measures, both Rescana guidance, untested against this bug, and not a substitute for Ghostscript 10.08.0 or your vendor's fixed package: run document conversion as an unprivileged user in an isolated container or sandbox with no outbound network access; and, where PostScript, EPS, PDF or XPS input is not needed, disable those ImageMagick or GraphicsMagick coders in the policy file so uploads are not handed to Ghostscript. No vendor we opened lists these as a formal workaround. Do not treat -dSAFER itself as a workaround: the CNA describes this CVE as a bypass of that restriction.

Indicators of Compromise

No source we opened publishes indicators of compromise for CVE-2026-125396: no hashes of malicious documents, no network addresses, no process names and no actor names. We do not publish any, and we found no public proof of concept in the sources we opened, so there is nothing to match against.

As Rescana guidance, treat unexplained Ghostscript or gs crashes, unexpected child processes of document converters or print servers, and new files written by the account that runs Ghostscript after a document was processed, as reasons to investigate that host and to confirm the Ghostscript version. We will update this advisory if the CNA, Artifex, CISA or a distribution publishes exploitation details or indicators.

Third-Party Risk (TPRM) Angle

Ghostscript is often an invisible dependency inside document conversion, preview, mail and print pipelines, including tools that accept uploads or attachments and hand them to ImageMagick, a print filter or a PDF renderer (Rescana's reading, consistent with the oss-security remark about other programs passing files to Ghostscript). A supplier that does not ship Ghostscript as a named product may still be running it. Rescana has no data on how widely the flaw is used or exploited.

Suppliers to ask first (Rescana guidance):

  • Document conversion, preview and e-signature vendors that accept PostScript, EPS, PDF or office uploads and render them server-side.
  • Mail, ticketing and content-management platforms, including WordPress hosts that enable Imagick.
  • Print-management and printer-fleet vendors whose servers or appliances render jobs with Ghostscript.
  • Linux distribution and appliance vendors that ship ghostscript in a base image.

Questions that give a usable answer: Do you invoke or embed Ghostscript, and which version? Do you process untrusted documents with it? Which release carries 10.08.0 or a backport of the fix, and when will it ship? Until then, what controls keep untrusted files away from the interpreter? For WordPress suppliers, have the GHSA-8vr3-7mxf-gx8w updates been applied as well as the interpreter update?

Need help working out which suppliers pass untrusted documents to Ghostscript with access to your data? Our team can help you prioritize them and track their answers. Talk to an expert

Sources

  • oss-security: Ghostscript 10.08.0 fixes multiple CVEs, including "upload2shell": Alan Coopersmith's announcement; 10.08.0 dated 2026-09-08; the other CVEs listed as fixed in that release; its report that research published the same day identified CVE-2026-125396 as also fixed in 10.08.0 and described ways it was exploitable through other programs passing files to Ghostscript; sent 14:09 EST, November 21, 2026; checked 06:27 EST, November 22, 2026.
  • oss-security follow-up: Jan Schaumann's reply, 14:59 EST, November 21, 2026, pointing to the August 12 WordPress advisory and arguing the issue should be treated as public from the September 8 fix date; checked 06:29 EST, November 22, 2026.
  • VulnCheck advisory and the CVE record for CVE-2026-125396, read through the CVE API: CNA VulnCheck; reserved 17:03 EST and published 18:28 EST, November 9, 2026; CWE-648; CVSS 4.0 8.5 and CVSS 3.1 7.8; affected versions; credits; last updated 23:06 EST, November 21, 2026; checked 06:27 EST, November 22, 2026.
  • NVD API 2.0: CVE-2026-125396: status Awaiting Analysis; published 19:17 EST, November 9, 2026; last modified 00:18 EST, November 22, 2026; VulnCheck scores as Secondary; CISA Coordinator SSVC at 11:29 EST, November 10, 2026; checked 06:27 EST and rechecked 06:48 EST, November 22, 2026.
  • GitHub Advisory Database: GHSA-hg3g-c775-3fpj: unreviewed entry, severity high, CVSS 4.0 8.5 and CVSS 3.1 7.8, published 19:31 EST, November 9, 2026, updated 01:30 EST, November 22, 2026; checked 06:27 EST, November 22, 2026.
  • Ghostscript News, current and the 10.08.0 docs News page: current page dates 10.08.0 as 2026-09-08 and says the release "addresses a number of potential security issues"; the versioned 10.08.0 page still shows 2026-05-19; checked 06:27 and 06:31 EST, November 22, 2026.
  • Ghostscript 10.07.0 News: 10.07.0 dated 2026-03-16; checked 06:30 EST, November 22, 2026.
  • Ghostscript CVE index: lists CVE-2026-110133, CVE-2026-112922, CVE-2026-103226, CVE-2026-19547 and CVE-2026-39919 as fixed in 10.08.0, and does not list CVE-2026-125396; retrieved 06:49 EST, November 22, 2026 (the same URL had not been retrievable at 06:27 EST).
  • Fix commit 1484854: commit message and dates; one file, Resource/Init/gs_fonts.ps; checked 06:29 EST, November 22, 2026.
  • Debian security tracker: CVE-2026-125396, with DSA-6516-1 and DLA-4914-1: fixed and vulnerable package versions; checked 06:27 and 06:29 EST, November 22, 2026.
  • Ubuntu CVE page: CVE-2026-125396: Ubuntu Medium priority, all listed releases Needs evaluation, last updated November 18, 2026; checked 06:27 EST, November 22, 2026.
  • SUSE CVE page: CVE-2026-125396: state Pending, important, CVSS 3.1 7.8 and CVSS 4.0 8.5, Bugzilla 1289822, Tumbleweed released package, SLE packages in QA; checked 06:28 EST, November 22, 2026.
  • Red Hat Bugzilla 2569593 (API): alias CVE-2026-125396, status New, severity high, no fixed-in version; checked 06:31 EST and rechecked 06:49 EST, November 22, 2026. The Customer Portal CVE page and data API still returned HTTP 404 at that recheck (those 404 URLs are not cited as links).
  • WordPress advisory GHSA-8vr3-7mxf-gx8w: CVE-2026-65640, High 8.8, published August 12, 2026, official description, affected and patched versions; checked 06:31 EST, November 22, 2026.
  • ENISA EUVD: EUVD-2026-120559 (facts taken from the EUVD API): assigner VulnCheck, CVSS 4.0 8.5, published 18:28 EST, November 9, 2026; checked 06:27 EST, November 22, 2026.
  • CISA: KEV catalog JSON feed: catalogVersion 2026.11.20, released 12:48 EST, November 20, 2026, 1,790 entries; CVE-2026-125396 not present; checked 06:27 EST and rechecked 06:48 EST, November 22, 2026.
  • FIRST: EPSS API, CVE-2026-125396: 0.00164, percentile 0.05606, date 2026-11-21; checked 06:27 EST, November 22, 2026.

Talk to an expert

Talk to Rescana about this advisory — or about how we help TPRM teams keep reviews moving.