~/f4n6 $ grep -r "OpenSSH auth bypass CVE-2026-123158 added to CISA KEV" ./investigations/ --include="*.md"

OpenSSH auth bypass CVE-2026-123158 added to CISA KEV

Jeff Davies 09 Nov 2026 6 min read

1. Executive summary

CVE-2026-123158 is an authentication bypass in OpenBSD OpenSSH on Linux that lets a remote attacker log into passwordless accounts without knowing a password. NVD records it at CVSS 8.1 (HIGH) and CISA added it to the KEV catalog on 2026-11-09, with a remediation due date of 2026-11-12. The flaw is exploited in the wild: Previdian reported eight attempts from a single Japan-based address and confirmed successful root access on a honeypot running real Alpine Linux and OpenSSH, less than nine hours after CVE publication. It affects only OpenSSH before 10.6p2 on musl-based Linux without PAM password checks, so Alpine Linux, OpenWrt, Buildroot, Yocto and many container images are in scope; glibc and PAM-checking deployments are not. For EMEA financial services the exposure is concentrated in container base images and musl-based edge or embedded builds that run sshd with password logins enabled, and the fix (OpenSSH 10.6p2) is available.

2. Regulatory framing

Article Trigger (the fact in this item) Practical impact
DORA Art. 28: ICT third-party risk — general principles The affected component is an open-source SSH server shipped inside Alpine, OpenWrt, Buildroot, Yocto and container base images that clients pull from third parties. Clients must establish whether sshd in their estate comes from a third-party image or distribution build and hold that supplier to a patched version (10.6p2).
DORA Art. 30: key contractual provisions with ICT third-party providers Where the vulnerable sshd is supplied through a managed service or image provider, the patch obligation sits with that provider. Contract terms should require the provider to ship 10.6p2 or document an equivalent mitigation, and to notify on exploited-component updates.

3. Technical analysis & attack chain

The vulnerability is a logic flaw in how sshd handles accounts that carry no usable password hash. The confirmed exploitation sequence, as reported by Previdian and the finder Özgün Kültekin, is:

  1. The attacker reaches an exposed SSH port on a musl-based Linux host running OpenSSH before 10.6p2, with password authentication enabled (the OpenSSH default).
  2. The target account has a * in place of a password hash in /etc/shadow. Root has shipped with * since around Alpine 3.16, and key-only users created without a password and users pulled from a directory service through musl-nscd also fit.
  3. The attacker submits a password longer than 256 characters. Since 2012 musl's crypt() refuses SHA-256 and SHA-512 passwords longer than 256 bytes and returns * on failure. That output matches the stored *, and sshd authenticates the session.
  4. Previdian observed the source IP 89.116.89[.]18 use oversized passwords to bypass authentication and run reconnaissance commands (id, uname, hostname) through non-interactive SSH sessions.

The mechanism traces to a 2016 OpenSSH change made to close a username-enumeration timing issue (CVE-2016-6210): when an account has no usable hash, sshd borrows another account's salt and hashes the password anyway. On musl that hash attempt fails and returns *, which collides with the stored *. Kültekin tested builds across combinations and threw roughly 17,600 unusual login attempts at them; only the musl build without PAM accepted logins it should have refused.

Preconditions are specific. The host must use musl and not check passwords through PAM, password logins must be enabled, and there must be a target account with only * in its password field plus another local account that has a real password hash. Alpine's default openssh-server package is built this way. Kültekin also lists OpenWrt, Buildroot and Yocto, and notes many pulled Docker images. On Alpine's official AWS and Azure images the default alpine user is passwordless and has passwordless doas, so a bypass of that account yields root; SOCRadar called indirect root access the larger concern. OpenSSH's default PermitRootLogin setting keeps direct root login out of reach unless an administrator has changed it.

The fix is OpenSSH 10.6p2, released 2026-11-05. The Alpine Linux team's advisory states the attack takes one connection, works every time, and needs no retries.

Attribution and tooling: no actor is named in the reporting. Previdian's Ryan Dewhurst attributed the speed of exploitation to attackers using AI, which is his assessment, not a finding about a specific group. The finder stated he used "agent swarms" during research; that concerns the research, not the attacks. The exploitation detail (eight attempts, one sensor, one source IP, confirmed root on a honeypot) is single-sourced to Previdian and should be treated as such.

4. Mitigation & containment

P1 (within 24 hours)

  • Upgrade OpenSSH to 10.6p2 on every musl-based Linux host that runs sshd with password authentication. This is the vendor fix and closes the flaw.
  • Where 10.6p2 cannot be deployed immediately, disable password authentication in sshd_config (PasswordAuthentication no) so the bypass path is not reachable, and confirm key-based access still works.
  • Inventory container base images and embedded builds (Alpine, OpenWrt, Buildroot, Yocto) for sshd versions before 10.6p2 and rebuild or repin them to a 10.6p2 base.
  • Block or alert on inbound SSH authentication attempts carrying passwords longer than 256 characters at the perimeter and on host logs.

P2 (within 72 hours)

  • Audit /etc/shadow for accounts with * in the password field that are reachable over SSH, and remove SSH access for any that do not need it. On Alpine cloud images, review the default alpine user and its passwordless doas rights.
  • Check whether PermitRootLogin has been changed from the default on any musl host; if it has, revert it unless there is a documented need.
  • Hunt SSH logs for the reconnaissance commands Previdian observed (id, uname, hostname) run through non-interactive sessions from unexpected sources.

P3 (within 7 days)

  • Add image-build checks that fail on OpenSSH versions before 10.6p2 in musl-based images.
  • Review third-party image and managed-service contracts against the DORA Art. 28 and Art. 30 points in §2.

No fixed release other than OpenSSH 10.6p2 is named in public reporting. No vendor patch identifier beyond the version is given.

5. Indicators of compromise

type value confidence source
ipv4 89.116.89[.]18 medium (single-sourced) Previdian via Help Net Security
ipv4  89.116.89[.]18

The IP is single-sourced to Previdian and reflects one sensor's observations; verify before blocking at scale.

Behavioural indicators

behaviour where to observe confidence
SSH authentication succeeds with a password longer than 256 characters sshd auth logs, PAM/sshd debug logs medium (single-sourced)
Non-interactive SSH sessions running id, uname, hostname in sequence process accounting, EDR process telemetry, shell history medium (single-sourced)
Successful login to an account whose /etc/shadow password field is * sshd logs correlated with /etc/shadow medium

6. Detection

title: OpenSSH oversized-password authentication bypass attempt
id: 8f3c2a10-6b7e-4d21-9a55-0c1e2f7b9d44
status: experimental
description: Detects SSH authentication where an oversized password is submitted against a passwordless account, consistent with CVE-2026-123158 exploitation.
author: Adverse Trace
date: 2026-11-09
references:

  - https://www.helpnetsecurity.com/2026/11/06/openssh-sshole-cve-2026-123158-exploitation/
logsource:
  product: linux
  service: sshd
detection:
  selection:
    message|contains:

      - 'Accepted password'
  condition: selection
falsepositives:

  - Legitimate long passphrase logins on non-musl systems
level: high

The Sigma rule above is a starting point only: sshd does not log password length by default, so pair it with host-side auditing of authentication payload size or with the behavioural indicators in §5. No file artefacts, hashes, mutexes or command strings beyond the reconnaissance commands are given in the sources, so no YARA rule is emitted.

CVE assessment

1 referenced CVE — 1 actively exploited (CISA KEV)

CVE CVSS Exploited EPSS Summary
CVE-2026-123158 8.1 High ⚠ KEV 2026-11-09 1% sshd in OpenSSH before 10.6p2 on Linux, when musl libc is used without PAM, lacks a workaround for a password authentication by…

7. Sources

  • NVD, CVE-2026-123158 detail, https://nvd.nist.gov/vuln/detail/CVE-2026-123158
  • Help Net Security, "OpenSSH flaw exploited in the wild just hours after disclosure (CVE-2026-123158)", https://www.helpnetsecurity.com/2026/11/06/openssh-sshole-cve-2026-123158-exploitation/, 2026-11-06

8. Adverse Trace position

Adverse Trace assesses CVE-2026-123158 as HIGH (CVSS 8.1) and confirmed exploited in the wild, with CISA KEV listing on 2026-11-09 and a remediation due date of 2026-11-12. Client impact for EMEA financial services is real but bounded: exposure requires musl-based Linux without PAM, password logins enabled, and a passwordless account reachable over SSH, which points at container base images and embedded or edge builds rather than mainstream glibc server fleets. The exploitation evidence is single-sourced to Previdian and we treat the source IP and attempt count as indicative, not exhaustive. We will track for additional exploitation reporting and for confirmation of affected/fixed versions beyond OpenSSH 10.6p2, and we will reissue if a second independent source corroborates the in-the-wild activity or names an actor. Clients should prioritise upgrading musl-based sshd to 10.6p2 and disabling password authentication where they cannot.


Read the original source →

Published via PulseTrace — Adverse Trace threat intelligence.

Post this to LinkedIn
Formatting is converted automatically — headings, bullets, a link back & hashtags. Paste straight in.
J
Jeff Davies