CYBERRAKSHAK LABS · RESEARCH #087

🔐 OpenSSH “SSHole” Authentication Bypass

CVE-2026-123158 | Remote authentication bypass on affected musl-based Linux systems

By Vivek Kumar · Published 9 November 2026
RESEARCH#087
CATEGORYVulnerability Analysis / Linux / OpenSSH / Authentication Bypass
CRL ASSESSMENTVULNERABILITY ANALYSIS
RESEARCH LEVELTechnical Research
PUBLISHED2026-11-09
Source & social links:
LinkedIn Post ↗ WhatsApp ↗ YouTube ↗
How CyberRakshakLabs researches threats →
SSHole (CVE-2026-123158) is not a universal OpenSSH bypass. It affects a narrow combination of musl libc, a non-PAM sshd authentication path and specific password-field conditions. This research separates vendor observations from technical assessment and treats exploitation claims with care.
Scope and source note:

This report follows the supplied technical research document. Patch status is a dated snapshot as of 9 November 2026; verify current distribution advisories before operational action. Honeypot observations are attributed to their reporting sources and are not treated as proof of a named threat actor.

Executive summary

OpenSSH 10.6p2, released on 5 November 2026, fixes a password-authentication bypass nicknamed SSHole and tracked as CVE-2026-123158. The flaw affects a narrow configuration: OpenSSH running on Linux with musl libc and PAM disabled. Under additional account conditions, a remote attacker able to reach sshd can submit an unusually long password and authenticate to an account whose /etc/shadow password field is the asterisk (*) marker used for accounts without a usable password hash.

This is not a universal OpenSSH bypass. It does not authenticate to accounts with valid password hashes. Alpine Linux is a prominent affected example because its standard OpenSSH server build uses musl without PAM. Other musl-based systems, appliances and container images should be assessed individually.

Exploitation evidence must be described carefully. Previdian reported a successful root login on its Alpine honeypot. Separately, SANS ISC observed long-password scans and clear exploit attempts in Cowrie honeypots, but its 8 November diary did not confirm successful sessions on its own sensors.

1. Vulnerability profile and affected conditions
  • CVE: CVE-2026-123158; nickname: SSHole / *sshole.
  • Component: OpenSSH server (sshd) password-authentication path.
  • Vulnerable platform: Linux using musl libc with PAM disabled for the sshd authentication path.
  • Fixed upstream version: Portable OpenSSH 10.6p2 or later; distribution backports may use different package version strings.
  • CWE: CWE-394, Unexpected Status Code or Return Value.
  • Severity: MITRE/NVD CVSS 3.1 score 8.1 High. Amazon Linux's advisory reports a 9.8 assessment. Refer to the applicable vendor/CNA score rather than presenting a single universal score.
  • Successful exploitation requires password authentication enabled, a target account whose stored password field is '*', and another local account with a valid SHA-256 or SHA-512 crypt hash ($5$ or $6$), in addition to network reachability and the musl/no-PAM condition.
2. Technical root cause

OpenSSH verifies a password by hashing the submitted value through the platform crypt() implementation and comparing the result with the account's stored password field. Accounts with no usable password hash commonly store '*' in /etc/shadow; this marker should not match a valid password.

musl crypt() returns the literal '*' on certain errors. Its SHA-crypt implementation rejects password inputs longer than 256 bytes and returns '*' as an error value. That string is not a valid hash; it signals that hashing failed.

OpenSSH's mitigation for CVE-2016-6210 (a username-enumeration timing issue) includes a path that borrows salt/hash parameters from another account when the target account has no usable hash, so password-check timing does not reveal whether the account exists. In the affected non-PAM musl build, the oversized password makes crypt() return '*'. sshd can then compare that error marker with the target account's stored '*' marker and incorrectly accept authentication.

The root cause is the interaction between musl's error-return convention and sshd's password-hash comparison path. OpenSSH 10.6p2 adds a workaround in sshd. OpenSSH's own release note says the root cause is outside sshd but exploitable through sshd and avoidable there.

3. Attack chain and potential attacker objectives
  1. Discover exposed SSH services and identify likely musl-based Linux targets, such as Alpine hosts, embedded devices or containers that run sshd.
  2. Identify a candidate username for an account with a '*' password marker. Candidate accounts may include key-only deployment users, directory-backed users or cloud-image default users.
  3. Connect to the SSH service and submit an oversized password in the documented range (commonly described as 257–1,024 bytes). No valid password or SSH key for the target account is needed if all preconditions are met.
  4. If authentication succeeds, obtain a session under that account's existing privileges. The bypass does not automatically grant root unless the target account already has root-equivalent privileges or the attacker can escalate.
  5. Enumerate host identity, groups, sudo/doas permissions, services, accessible files, environment variables and mounted secrets.
  6. Seek deployment credentials, API tokens, cloud credentials, CI/CD secrets, private keys, application/database passwords and other secrets available to the compromised account.
  7. Use accessible credentials or trust relationships for lateral movement, persistence, data theft, application modification or service disruption. These are plausible post-access actions, not claims that every observed attempt performed them.
4. Platforms and exclusions

Alpine Linux is a principal affected distribution. OpenWrt, Buildroot, Yocto and other musl-based systems merit review when they ship OpenSSH built without PAM. Container images and cloud images should be included in the inventory if they run sshd.

Standard glibc-based Debian, Ubuntu and Fedora builds are not affected by this specific musl error-return path. Red Hat states its products are not affected because its OpenSSH builds use glibc and PAM by default. PAM-enabled configurations using the supported PAM path are outside the described vulnerable condition. Dropbear and OpenBSD's native OpenSSH builds are not affected by this specific issue.

Do not infer that every Alpine container is exposed. Confirm that it runs sshd, determine its OpenSSH build/package, check PAM and effective sshd settings, and verify the relevant account database conditions.

5. Threat actor assessment and exploitation observations

The public reporting reviewed by 9 November does not attribute SSHole exploitation to a named APT, ransomware group or state. Rapid exploitation after disclosure is not evidence of a specific operator identity.

Previdian reported its first observed attempt less than nine hours after CVE publication and stated that it confirmed root access on an Alpine Linux/OpenSSH honeypot. It identified source IP 89.116.89[.]18 and reported non-interactive reconnaissance commands including id, uname and hostname. Initial coverage described eight attempts from that single Japan-geolocated address. This is a honeypot vendor's observation and was not independently verified by the news outlets.

SANS ISC's Cowrie sensors recorded unusually long password requests: 23 on 1 November, 5 on 3 November, 58 on 4 November, 232 on 5 November, 713 on 6 November, 246 on 7 November and 51 on 8 November. SANS identified its first clear SSHole exploit attempt on 5 November. It said 1 November inputs appeared to be unrelated buggy brute-force scripts and noted its sensors capture only the first 255 password characters. SANS had not confirmed successful sessions in its published 8 November diary.

IP addresses in honeypot data are infrastructure indicators, not actor identities. They may represent compromised hosts, proxies, rented servers or scanning infrastructure. Do not infer nationality or attribution from geolocation or ASN.

6. Historical context and timeline

2012: musl crypt() behaviour described in the technical disclosure was introduced; overlong SHA-crypt keys can return '*' as an error value.

2016: OpenSSH changed password-check handling in response to CVE-2016-6210 to reduce username-enumeration timing differences. The interaction with musl's error return created the conditions for SSHole.

1 November 2026: SANS sensors began seeing unusually long password inputs, initially considered likely unrelated buggy brute-force scripts.

5 November 2026: CVE-2026-123158 was published and OpenSSH 10.6p2 released. SANS observed its first clear exploit attempts using long all-'A' passwords. Previdian later reported an attempt less than nine hours after publication.

6 November 2026: Help Net Security and The Hacker News reported Previdian's claim of successful root access on an Alpine honeypot. SANS recorded 713 long-password requests in its honeypots.

7 November 2026: Korea's CERT published an advisory describing the musl/PAM-disabled condition and recommending prompt updates.

8 November 2026: SANS published its scan telemetry and initial scanner IP counts; its diary did not confirm successful exploitation on its own honeypots.

9 November 2026: Cybersecurity News published the supplied report summarizing SANS findings and recommending OpenSSH patching and sshd restart.

7. Detection and threat hunting

Prioritize authentication-log review for successful password logins to accounts expected to be key-only or passwordless. Alpine recommends looking for 'Accepted password for' events involving accounts such as alpine, root (where password login should be prohibited) or key-only users.

grep -E 'Accepted password for|session opened' /var/log/messages
journalctl -u sshd --since '2026-11-01' | grep -E 'Accepted password|session opened'

Adapt log paths and service names to the distribution. Correlate source addresses, timestamps, successful password authentication, non-interactive sessions, and follow-on commands such as id, uname, hostname, doas/sudo, user creation or authorized_keys changes.

Review doas/sudo logs, new users/groups, ~/.ssh/authorized_keys and /root/.ssh/authorized_keys, systemd services, cron jobs, modified binaries, package-manager history, outbound connections and secrets accessible to the account.

Review account password markers and determine whether the required $5$/$6$ hash condition exists using secure administrative procedures. Never disclose /etc/shadow contents in public tickets or chats.

Network monitoring for unusually long SSH passwords may be limited because server logs typically do not record the submitted password. Honeypots can provide additional telemetry, but truncation and sensor design affect visibility.

8. Mitigation and remediation

Immediate temporary mitigation: disable password-based SSH authentication where operationally possible. Set PasswordAuthentication no, validate sshd configuration, ensure working key-based or console access, and restart sshd. This blocks the vulnerable password-authentication path.

Restrict SSH ingress to trusted management networks, VPNs or approved bastion hosts. Do not expose management SSH broadly to the internet.

Permanent remediation: install Portable OpenSSH 10.6p2 or the distribution's patched package/backport. Check the vendor advisory and package changelog rather than relying only on the upstream version banner.

Restart sshd after the update. Installing a package does not replace the already-running daemon until the service is restarted.

Inventory Alpine branches, OpenWrt/Buildroot/Yocto devices, container images, cloud images, appliances and CI runners that include an SSH server. Validate libc, PAM and effective authentication settings.

If a successful unexplained login to a passwordless account is found, treat it as a potential compromise. Preserve logs, rotate credentials and tokens accessible to the account from a clean system, revoke exposed SSH keys, investigate lateral movement and rebuild if unauthorized root access or persistence cannot be excluded.

9. Conclusion

SSHole is a configuration-dependent authentication bypass with potentially severe consequences for exposed musl-based Linux systems. Its importance is not that all OpenSSH servers are vulnerable, but that a remote attacker may enter accounts intentionally configured without passwords when the exact library, PAM and account conditions align.

Organizations should inventory musl-based SSH servers, patch to OpenSSH 10.6p2 or vendor-fixed packages, restart sshd, disable password authentication where patching is delayed, and investigate successful password logins to passwordless accounts. Attribution remains unknown; honeypot IPs and observations should be treated as indicators, not proof of a named threat actor.

References