gpg.fail retrospective RCE claim in gpgsm 2.4.9 disputed by GnuPG's Werner Koch: only a segfault reproduced
The gpg.fail authors' retrospective disclosed an unpatched RCE in gpgsm 2.4.9 via malicious certificate import with --debug options, but GnuPG maintainer Werner Koch says testing reproduced only a segfault from a DER-as-printf format string flaw in --debug…
On 2026-09-13, Sam James shared on oss-security that the researchers behind the 'gpg.fail' set of GnuPG vulnerabilities published a retrospective including a talk recording and slides, with a proof-of-concept in their repository; the slides appeared to hint at an additional vulnerability not yet widely reviewed. In a same-day reply, Peter Gutmann referenced a Hacker News discussion of the retrospective, arguing that a recurring source of severe PGP vulnerabilities is the format's permissive packet system, in which a PGP message is a practically arbitrary stream of packets; that reply disclosed no new vulnerability or CVE. On 2026-09-14, the retrospective's contents were clarified: it revealed a remote code execution vulnerability in gpgsm 2.4.9, triggered when gpgsm is invoked as "gpgsm --debug all --import" with an attacker-controlled malicious certificate file; the researchers never reported the issue to GnuPG, leaving it unpatched as an apparent 0-day, though exposure appears limited because exploitation requires importing untrusted certificates with debug options enabled. On 2026-09-15, GnuPG's Werner Koch disputed the RCE claim: GnuPG versions above 2.2 produce garbled stderr or crash when certificates are used with --debug x509 because DER data is passed as a printf format string; testing the researchers' repo certificates yielded only a segfault, not code execution, and Koch states it is unclear how remote code execution would be achieved, asking for a real reproducer. No CVE identifier was given in any of the reports, and the sources disagree on whether the flaw is an exploitable RCE or merely a format-string-induced crash.
- gpg.fail authors published a retrospective with talk recording, slides, and a repository PoC; slides hint at another vulnerability not yet widely reviewed
- Claimed RCE affects gpgsm 2.4.9, triggered by "gpgsm --debug all --import" with an attacker-controlled malicious certificate file
- Issue was never reported to GnuPG; no patch available, leaving an apparent 0-day; exposure appears limited since it requires importing untrusted certificates with debug options enabled
- Werner Koch (GnuPG) says GnuPG versions > 2.2 pass DER certificate data as a printf format string in the --debug x509 path, causing garbled stderr or a crash
- Koch's testing of the researchers' Git repo certificates reproduced only a segfault, not code execution; he calls the RCE mechanism unclear and requests a real reproducer
- Peter Gutmann's reply attributes recurring severe PGP vulnerabilities to the format's permissive packet system (messages as practically arbitrary packet streams); it disclosed no new vulnerability or CVE
- Sources disagree on severity: researchers claim RCE, GnuPG maintainer has reproduced only a crash; no CVE identifier was given in any report
Coverage timelineoldest first · each row is one article
- · 2d agoRetrospective by 'gpg.fail' authors
oss-security· 52
Authors of the gpg.fail GnuPG vulnerability set published a retrospective with talk recording, slides, and a repo PoC.
- · 2d agoRe: Retrospective by 'gpg.fail' authors
oss-security· 15
Peter Gutmann replies to gpg.fail retrospective, pointing to PGP's arbitrary packet format as root of many past vulnerabilities.
- · 1d agoRe: Retrospective by 'gpg.fail' authors
oss-security· 48
Unreported 0-day RCE in gpgsm 2.4.9 disclosed via gpg.fail talk, triggered by importing a malicious certificate with --debug options.
- · 6h agoRe: Retrospective by 'gpg.fail' authors
oss-security· 20
GnuPG's Werner Koch says gpg.fail samples only crash GnuPG via DER-as-printf format string in --debug x509; RCE claim remains unproven.