A Four-Year-Old OpenSSH Fix Just Opened a Password-Free Login on Alpine Linux
Cybersecurity

A Four-Year-Old OpenSSH Fix Just Opened a Password-Free Login on Alpine Linux

CVE-2026-123158 lets attackers log into passwordless accounts on musl-based Linux systems with no PAM, and honeypots clocked exploitation within hours of disclosure.

PublishedNovember 14, 2026
Read time5 min read
Share

A patch from a decade ago just backfired

CVE-2026-123158 is not a new feature flaw. It is the unintended consequence of a 2016 OpenSSH change that fixed a timing side channel attackers could use to enumerate valid usernames. That fix made sshd borrow the cryptographic salt from a real account's password hash whenever the target account's password field was too short to contain one, which is exactly what happens on accounts marked with a bare asterisk to indicate no password is set. For nearly a decade that behavior was harmless on the Linux distributions most enterprises run.

The second half of the bug sat in musl libc, the small C library Alpine Linux and a long list of embedded and container-focused distributions use in place of glibc. Since 2012, musl's crypt function has rejected any password longer than 256 bytes for SHA-256 and SHA-512 style hashes by returning the literal string asterisk instead of raising an error. On a PAM-free sshd installation, that returned asterisk matches the stored asterisk in the passwordless account's record, and the login succeeds. Put the two decade-old decisions together and you get a working, unauthenticated login bypass that nobody wrote on purpose.

Why Alpine specifically, and why it matters more than it sounds

Four conditions have to hold simultaneously: sshd runs without PAM, password authentication is enabled, at least one account has no password set, and at least one other local account has a real SHA-256 or SHA-512 hash to borrow a salt from. That is precisely the out-of-the-box state of Alpine's openssh-server package, which is why this reads as an Alpine bug even though the researcher, Özgün Kültekin of Wakefield, flagged that other musl-based builds including OpenWrt, Buildroot, and Yocto warrant the same check.

The part that should concern infrastructure teams specifically is the indirect privilege escalation path. Alpine's tiny-cloud cloud images between versions 3.18.11 and 3.24.3 ship a default alpine user that can run commands as root through doas without a password. An attacker who authenticates as that user through the sshole bypass does not need a second vulnerability to reach root. Given how widely Alpine is used as a minimal base image across container fleets, the practical exposure depends entirely on whether a given image bothers to run an SSH server at all, which most production container images do not, but build, CI, and bastion hosts frequently do.

The exploitation timeline is the real headline

MITRE published the CVE record on November 5. Honeypot operator Previdian recorded its first exploitation attempt less than nine hours later and counted 20 attempts from 13 distinct IP addresses by November 8. SANS honeypot telemetry showed SSH logins carrying unusually long passwords peaking at 713 instances on November 6. One researcher reported a successful compromise of a vulnerable sensor on November 7, where the attacker queried cloud metadata services and attempted to read root's SSH keys and the shadow password file through doas, without writing any files to disk, a pattern consistent with an attacker who already understood exactly what to look for on this specific misconfiguration.

CISA added the vulnerability to its Known Exploited Vulnerabilities catalog on November 9 and set a federal remediation deadline of November 12, a three day turnaround that reflects how quickly this moved from disclosure to confirmed exploitation. No public confirmation of a production compromise exists yet, but the honeypot data removes any argument that this is theoretical. Public proof of concept code is already circulating on code-sharing sites, unverified and possibly malicious itself, which lowers the bar for less sophisticated actors to pile on.

What OpenSSH maintainers are actually telling you

OpenSSH released version 10.6p2 as an emergency fix on November 5, changing sshd to reject a password match whenever the stored password field is shorter than four characters, closing the salt-borrowing path regardless of what the underlying libc does with overlong input. OpenSSH describes this explicitly as a workaround rather than a root-cause fix, because the real defect sits in how musl's crypt function silently swallows an error condition instead of surfacing it. That framing matters for CTOs managing a mixed fleet: patching sshd fixes the immediate hole, but any other code path that relies on musl's crypt behaving the way glibc's does may still be carrying a latent bug nobody has found yet.

OpenSSH maintainer Damien Miller's response to the disclosure speed is worth repeating to your own security leadership. He said that long embargo periods on vulnerability disclosure now look dangerous in their own right, because AI-assisted vulnerability research can independently rediscover the same class of bug within hours. That is a direct challenge to the traditional coordinated disclosure timeline most vendors still plan around, and it is a reasonable basis for revisiting how your organization prioritizes emergency patches that arrive with little or no advance warning.

The fix is simple, the audit is the hard part

Mitigation for confirmed Alpine systems is straightforward: run apk upgrade to pull the patched openssh package, then restart the sshd service, since installing the update alone does not restart a running daemon. Point releases 3.21.10, 3.22.8, 3.23.8, and 3.24.4 or later already ship the fix baked in. Where immediate patching is not possible, disabling password authentication entirely, ideally through a drop-in configuration file rather than editing sshd_config directly, closes the hole without requiring a restart of other services that depend on SSH.

The harder job is inventory. Most enterprises cannot answer quickly which of their container base images, CI runners, network appliances, and embedded Linux devices are built on musl rather than glibc, because that distinction has rarely mattered for security purposes before now. Teams should audit for Alpine or other musl-based images running an exposed SSH service, check whether PAM is active, and review authentication logs for accepted password logins against accounts that should never have had a password in the first place, including root, key-only service accounts, and the default alpine cloud-image user.

Tagged#news#security#cybersecurity#breach#cisa#ransomware#zero-day#supply-chain#ai-security#openssh#alpine-linux#musl-libc#authentication-bypass#container-security#linux-infrastructure#cve-2026-123158