GnuPG CVE-2026-24881

Exploitation requires gpg-agent to process an attacker-crafted CMS (S/MIME) EnvelopedData message through the PKDECRYPT --kem=CMS code path, meaning the target must use a GnuPG build before 2.5.17 that includes the KEM/CMS decryption feature and the victim must hold a private key the message is encrypted to and actually attempt decryption (interactively in a mail client or via an automated decryption service). Limiting factors: the vulnerable PKDECRYPT --kem=CMS handling is specific to the newer 2.5.x development branch, so 2.4.x stable production deployments may not be affected; exploitation depends on the victim invoking decryption of the malicious message rather than a passively listening network service; and reliable code execution (versus a crash) is constrained by stack protections (canaries, ASLR, NX) on the target platform.

Signals are mixed and should be weighed rather than taken at face value. The published CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H, 9.8) treats this as a trivially exploitable, unauthenticated network attack with full impact, and SSVC-style reasoning would flag high technical impact in a security-critical key-handling daemon. However, exploitation realistically requires a victim to attempt decryption of an attacker-supplied CMS message with a matching private key, which is closer to user interaction or automated mail-pipeline processing than to a directly reachable network listener - the NVD UI:N is arguably optimistic and should be verified against the real deployment. Reliable code execution from a stack overflow on modern targets with stack canaries, ASLR, and NX is non-trivial, so the dominant near-term outcome is denial of service (agent crash) with RCE as a credible but harder upside. EPSS is low (0.20%) despite public POC availability and the high CVSS, and the issue is absent from CISA KEV, suggesting that no broad opportunistic exploitation is occurring; treat this as a high-priority patch for systems that process S/MIME (mail gateways, automated decryption services) and a medium priority elsewhere.

An attacker emails a victim a malicious S/MIME EnvelopedData message whose wrapped session key field declares an oversized length; when the victim (or an automated mail-processing service) decrypts it, gpg-agent copies the attacker-controlled key data past a fixed stack buffer, crashing the agent and potentially corrupting the stack to gain control of execution within the key-handling process. Public proof-of-concept material is referenced from the GnuPG tracker (T8044), and because the trigger is a single crafted message, mass-mailing makes denial of service straightforward while code execution is a harder, target-specific escalation.