Koffei

CVE-2026-107441

🔴 High

Zyxel GS1900 series web-managed switches: GS1900-8, -8HP, -10HP, -16, -24, -24E, -24EP, -24HPv2, -48 and -48HPv2 · Each model at Patch 2 or earlier, that is 2.90(AAHH.2)C0, (AAHI.2), (AAZI.2), (AAHJ.2), (AAHL.2), (AAHK.2), (ABTO.2), (ABTP.2), (AAHN.2) and (ABTQ.2) C0 respectively. The record spells one model GS1900-HPv2 where the advisory says GS1900-24HPv2, which reads as a typo in the record rather than a separate product

CVSS (Koffei note)
8.8
CVSS (v3.1, CVE List)
8.8
EPSS (2026-11-18)
0.293%
22.20th percentile

Exploitation

Yes in substance, and two sources say otherwise, so both readings are here. CISA-ADP records exploitation none (SSVC 03/11/2026) and it is not in CISA KEV ; CERT Santé's alert of 03/11/2026 says it is not actively exploited and that no public proof of concept exists. Against that: Binarly found a working exploit for it on 17/09/2026 in an open directory on a server that was also acting as a command server for JITTERLY, and reported the flaw to Zyxel on 29/09/2026. The Hacker News of 05/11/2026 describes it as the flaw the live GS1900 exploit relies on, the one that makes the CVE-2026-7273 Patch 2 fix ineffective. This row follows the researchers who held the exploit over the trackers that had not caught up, which is why it is 🔴 while KEV and SSVC still read clean. Rechecked on 06/11/2026 and the trackers still have not moved: no KEV entry, SSVC still exploitation none, EPSS unchanged at 0.293%. But CISA-ADP touched the record again on 05/11/2026 at 04:55 UTC without changing the SSVC timestamp of 03/11/2026, so something in that container changed and the record does not say what. This is the known ADP pattern where providerMetadata.dateUpdated is the only reliable change signal, and it is worth re-reading before the next run Rechecked on 09/11/2026: the trackers have still not moved and the ADP container has not changed again. No KEV entry, SSVC still exploitation none timestamped 03/11/2026, EPSS unchanged, and the record still holds one reference and no exploit-tagged reference. The CISA-ADP container's dateUpdated is still 05/11/2026 04:55 UTC, so whatever changed then added no reference and no visible SSVC change, and nothing has changed since. The row keeps 🔴 on Binarly's evidence over the trackers

Timeline

  1. 03/11/2026CVE record published (CVE List)
  2. 05/11/2026Second Zyxel GS1900 flaw slips past the patch for an already-exploited bug
  3. 06/11/2026Zyxel's own patch doesn't close its GS1900 switch zero-day
  4. 09/11/2026Zyxel GS1900 switches zero-day, patch bypassed twice

Risk

8.8 High (v3.1, Zyxel CNA) — AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H, EPSS 0.293% (22.05th percentile, unchanged when rescored 08/11/2026). Three parties score this differently and one of them disagrees about root: CERT Santé gives the same 8.8 and the same vector but adds that exploitation does not grant privileged rights, while Binarly, who found it, gives 9.8 and says an unauthenticated remote attacker executes arbitrary OS commands with root privileges (CWE-787 out-of-bounds write in the CGI program, which Binarly identifies as dispatcher.cgi)

Fix

Yes — Patch 3, posted 23/10/2026 (Zyxel advisory of 03/11/2026): 2.90(AAHH.3)C0, (AAHI.3), (AAZI.3), (AAHJ.3), (AAHL.3), (AAHK.3), (ABTO.3), (ABTP.3), (AAHN.3) and (ABTQ.3) C0 for the ten models in order. Patch 3 checks the parameter counter before the copy and limits the copy size, which is what Patch 2 failed to do. No workaround exists. If a switch sat on Patch 2 or earlier while reachable, treat the firmware update as necessary but not sufficient and check credentials and configuration as well

Official description

An out-of-bounds write vulnerability in the CGI program of Zyxel GS1900-24E firmware versions through 2.90(AAHK.2)C0 could allow an unauthenticated adjacent attacker to exploit the flaw and potentially execute OS commands via a crafted HTTP request.

CVSS:3.1/AV:A/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H

CWE-787