Research

GPGME CVE-2026-129975: A Malicious OpenPGP Key Can Trigger a Double Free in the Library Your Mail Client Uses. Update to 2.3.1, Then Find What Links It

GPGME CVE-2026-129975: A Malicious OpenPGP Key Can Trigger a Double Free in the Library Your Mail Client Uses. Update to 2.3.1, Then Find What Links It

Update GPGME to 2.3.1 (or 1.24.4 if you are on the legacy branch) on every machine that runs a mail client, an XMPP client, Samba, Flatpak or Gpg4win, then restart those programs. On Windows that means Gpg4win 5.1.2. On Debian, install DSA-6646-1 (trixie) or DLA-4924-1 (bookworm) today. Programs that bundle or statically link their own copy of GPGME are not fixed by the system update: they need a rebuilt release from their vendor.

The GnuPG project published the advisory on Friday, November 20, 2026. Its wording: a crafted OpenPGP key with a very long user ID triggers a double free, which "can be used to exfiltrate data from applications," and "Exploiting this bug to mount an Remote Code Execution attack (RCE) can't be ruled out." It recommends an immediate update. No exploitation has been reported.

Am I affected?

Find the installed version first. Distributions backport fixes, so compare against your distribution's fixed package version below, not only the upstream number.

  • Debian or Ubuntu: run dpkg -l | grep -i gpgme and read the version column of the libgpgme package. Debian fixed versions: bookworm 1.18.0-3+deb12u1, trixie 1.24.2-3+deb13u1, sid 2.3.1-1.
  • Fedora, RHEL, Rocky, Alma: run rpm -q gpgme.
  • Arch: run pacman -Q gpgme.
  • macOS with Homebrew: run brew list --versions gpgme.
  • From source or unsure: run pkg-config --modversion gpgme. Older installs also have gpgme-config --version.
  • Windows: open Gpg4win's About dialog. You want 5.1.2 or later.

Then find what on the box uses it.

  • Debian or Ubuntu: apt-cache rdepends --installed on your libgpgme package name (for example libgpgme11t64 on trixie, libgpgme11 on bookworm) lists installed packages that depend on it.
  • RPM systems: rpm -q --whatrequires gpgme, and also rpm -q --whatrequires on the library name shown by rpm -ql gpgme.
  • Arch: pacman -Qi gpgme and read the "Required By" line.
  • A specific program: ldd /path/to/program | grep gpgme. A line naming libgpgme means it uses the shared system library and the system update covers it after a restart. No line does not mean safe: the program may carry a static or bundled copy, which only its vendor can fix.
  • After updating: programs that were already running keep the old library loaded until restarted. On Debian and Ubuntu, needrestart reports them. Simplest is a reboot or a logout and login for desktop mail clients.

Our edge sensors cannot see this: it is a bug in a local library reached through key and signature data, not an HTTP request, so we make no claim either way about scanning.

What is confirmed

  • Identifier: CVE-2026-129975, GnuPG tracker task T8557.
  • Affected: "All software using GPGME before version 2.3.0," whether as a shared library, a Windows DLL, or statically linked, per the advisory. The reporter confirmed it in current master and in "1.18.0 and later" as shipped by distributions. The advisory notes 2.3.0 was never publicly available.
  • Fixed: GPGME 2.3.1, dated 2026-11-19. On the legacy branch, the GPGME 1.24.4 release notes (2026-11-20) list "Fix a double free in the status line read code. T8557,[CVE-2026-129975]". The T8557 ticket itself says "1.24.3"; the release notes and the 1.24.3 date (May 2025) point to 1.24.4 as the right number. The project says the legacy branch "can't guarantee that it will see all security patches," so move to 2.3.1 where you can.
  • Gpg4win: 5.1.2 includes the fixed GPGME.
  • Not needing an update: the advisory says there is no need to update GPGMEPP or GPGMEQT, the C++ and Qt wrappers. They still sit on top of the C library, so the C library update is what matters for KDE programs.
  • The bug: in read_status() and read_colon_line() in src/engine-gpg.c, the code GPGME uses to read gpg's status lines and key listings. Both grow a buffer with realloc() and write the new pointer back only at the end. If a callback returns an error in between, the engine keeps a pointer to memory already freed, and frees it again when the engine is released.
  • The trigger, per the advisory: an OpenPGP key with a crafted, very long user ID. Its example of exfiltration is "by sending signed data." It does not list which user actions (importing a key, listing keys, verifying a signed message) reach the bug in each program.
  • Software the advisory names as affected: Gpg4win, GnuPG Desktop, GnuPG VS-Desktop, "most mail clients," "many XMPP clients," Samba, and Flatpak.
  • Workaround: none other than updating. For versions that cannot move to 2.3.1, the advisory offers a patch it says likely applies cleanly even to very old versions.
  • Credit and timeline: found by Johan Carlsson, reported to GnuPG by GitLab on November 12 (the ticket shows November 13). Disclosure was set for November 25 and moved up to November 20 because GitLab had to update earlier for the Ruby2Shell bug.

Distribution status, as of 17:30 UTC Saturday, November 21

  • Debian: fixed. DSA-6646-1 (November 20) ships gpgme1.0 1.24.2-3+deb13u1 for trixie. DLA-4924-1 (November 20) ships 1.18.0-3+deb12u1 for bookworm. Sid has 2.3.1-1. Forky (testing, 2.2.0-2) is still listed vulnerable.
  • Ubuntu: no CVE page and no security notice yet. The Ubuntu CVE API returns "does not exist" for this ID.
  • Red Hat: no CVE page yet (404) and no Bugzilla entry under this CVE alias.
  • NVD and the CVE list: no record published yet. NVD returns zero results and cve.org returns "record does not exist." There is no CVSS score from any source we could read. "Critical" is the press headline; the advisory says "immediate update," and Gpg4win calls it severe.
  • CISA KEV: not listed (catalog version 2026.11.20).

If you run Ubuntu or RHEL, watch those trackers and install the gpgme update the day it lands. Until then, be careful importing keys from strangers in any GPGME-based client, though the advisory does not say avoiding imports is enough.

What is likely, not confirmed

The advisory names categories ("most mail clients"), not products. These are well-known GPGME users we expect to be exposed, but no primary or distribution advisory we read names them for this CVE:

  • Mail clients: KMail and Kleopatra (KDE, via the C++ wrapper over libgpgme), Mutt and NeoMutt when built with GPGME support, Evolution, Claws Mail's PGP plugins.
  • Package and image signing: Arch's pacman verifies package signatures with GPGME. Podman and other tools built on the containers image library can verify signatures through GPGME. OSTree, which Flatpak sits on, uses it as well.
  • Others: some password managers and Git front ends that call GPGME rather than the gpg command line.

Run the dependency checks above to turn this list into facts for your own machines.

What is not known

  • Which exact user actions in each named program reach the bug, and whether a signed message alone (with no key import) is enough in a given client.
  • Whether a working code execution exploit exists. The project says RCE "can't be ruled out," not that it was shown.
  • CVSS scores and fixed package versions from Ubuntu and Red Hat, which had not published at the time of writing.

What our data shows

  • Blog: no prior post on GPGME or this CVE. Our morning brief flagged it as a gap.
  • Indicators: none published in the sources we read. This is a library flaw with no attacker infrastructure, so we fed nothing. Our IOC index has no records for "GPGME" or CVE-2026-129975, and CISA KEV does not list it.
  • Edge: not applicable, as above. No first-party sighting, and none possible from our sensors for this kind of bug.

Sources

  • GnuPG security advisory, gnupg-announce, November 20, 2026: https://lists.gnupg.org/pipermail/gnupg-announce/2026q4/000517.html
  • GnuPG tracker T8557: https://static.dev.gnupg.org/T8557.html
  • GPGME legacy branch release notes, T7659 (1.24.4): https://dev.gnupg.org/T7659
  • Debian security tracker: https://security-tracker.debian.org/tracker/CVE-2026-129975 (DSA-6646-1, DLA-4924-1)
  • Gpg4win change history: https://www.gpg4win.org/change-history.html
  • The Hacker News, November 21, 2026, "GnuPG Warns Critical GPGME Flaw Could Expose App Data via Malicious OpenPGP Keys": https://thehackernews.com/2026/11/gnupg-warns-critical-gpgme-flaw-could.html
  • Checked with no record found: NVD API, cve.org, Ubuntu CVE tracker, Red Hat CVE pages and Bugzilla, CISA KEV.

Read next: another signing bug where the check you trusted did the opposite. The Signature Failed to Verify, So It Was Accepted: CVE-2026-15981


Every indicator in this post is in the feed. Free.

1.58M+ IOCs, STIX 2.1 / TAXII, 88% novel vs ThreatFox, exploited-CVE leads ahead of CISA. No credit card — a free API key in 30 seconds, and you can audit every claim above against the live endpoints.

Get your free key → analytics.dugganusa.com/stix/register