A 2016 Fix for One SSH Bug Just Created a Different One — Here's What CVE-2026-123158 Actually Does
Table of Contents
- What CVE-2026-123158 actually is
- Why "disabled password" accounts exist, and how this breaks that guarantee
- Who's actually exposed, and who genuinely isn't
- The exploitation timeline: from one honeypot hit to thirteen countries
- The decade-old fix that quietly set this up
- The fix, and what to check on your own systems right now
- Why this matters for anyone learning Linux administration
- What we're not claiming
Security researchers gave this one a blunt name — "SSHole" — and the mechanism underneath it is almost comically specific: an attacker who types a password long enough can get into an account that was deliberately set up to never accept any password at all. It's a narrow bug in scope, but where it does apply, the bypass is complete, unauthenticated, and already being tried against real servers.
What CVE-2026-123158 actually is
CVE-2026-123158, rated 8.1 on the CVSS scale, affects Linux systems running OpenSSH built against the musl C library with PAM (Pluggable Authentication Modules) disabled — a configuration that includes Alpine Linux's default OpenSSH server setup, according to The Hacker News' reporting on November 6, 2026 (updated November 8). The flaw lets an unauthenticated attacker supply an oversized password — specifically one between 257 and 1,024 bytes long — to gain access to accounts that were intentionally configured without a usable password.
Why "disabled password" accounts exist, and how this breaks that guarantee
On Linux, an account can have its password field in /etc/shadow set to a single asterisk (*), which signals "this account cannot log in with a password at all" — a deliberate lockout used for service accounts, key-only accounts, or any login meant to be reached only through SSH keys, never a typed password. The authentication check is supposed to compute a hash of whatever password is submitted and compare it against that stored value; since no real password hashes to a bare *, the comparison should always fail.
The bug breaks that assumption in a narrow but exploitable way: supplying a password of exactly the wrong length — 257 to 1,024 bytes — makes musl's own hash-computation step fail internally and return * as its result. That computed * then matches the account's stored *, and the login succeeds. Two conditions have to both be true for this to work: the target account's password hash has to already be the literal asterisk, and the system has to be running the vulnerable musl-plus-no-PAM combination. The Hacker News' reporting is explicit that this doesn't extend any further than that — it does not let an attacker bypass a real, working password hash on an otherwise normal account.
Who's actually exposed, and who genuinely isn't
This is a configuration-specific bug, not a universal OpenSSH flaw. Affected: Linux systems running OpenSSH compiled against musl with PAM disabled, which describes Alpine Linux's out-of-the-box sshd setup. Explicitly not affected, per the same reporting: Dropbear (a different, commonly used lightweight SSH server), any glibc-based Linux distribution (which covers Debian, Ubuntu, RHEL, and most mainstream distributions), non-Linux systems entirely, and any system where password checks actually go through PAM rather than around it. If your production fleet runs Ubuntu or RHEL with standard PAM-backed authentication, this specific CVE doesn't reach you — the exposure is concentrated in musl-based distributions like Alpine, which are common in minimal container base images precisely because they're small, which is its own reason to know whether any of your containers or edge devices are built on one.
The exploitation timeline: from one honeypot hit to thirteen countries
Security firm Previdian reported observing the first exploitation attempt less than nine hours after the CVE was published, and confirmed successful root access against a honeypot it had running Alpine Linux and OpenSSH. At initial publication, that tally stood at eight attempts from a single IP address geolocated to Japan. By the November 8 update, that had grown to 20 attempts from 13 distinct IP addresses spread across Denmark, Iran, Japan, the Netherlands, Norway, Singapore, Sweden, Switzerland, and the United States — a meaningful jump in both volume and geographic spread in just two days, and a reasonably strong signal that automated scanning for this specific condition is already running at some scale.
The decade-old fix that quietly set this up
The detail that makes this story more than just another CVE number: the underlying cause goes all the way back to a patch from 2016, originally written to close a completely different bug — CVE-2016-6210, a username-enumeration flaw that let an attacker tell valid usernames apart from invalid ones by measuring how long the server took to respond. That 2016 fix changed how failed lookups are handled internally, and a decade later, that exact change is the mechanism this new bug rides on. It's a useful, concrete example of something security-minded Linux administrators already know in the abstract: a fix for one problem can quietly create the conditions for a different one, years later, in a part of the system nobody was actively worried about.
The fix, and what to check on your own systems right now
OpenSSH addressed this in version 10.6p2. The practical step is installing that version, or your distribution's equivalently patched package, and restarting sshd. Alpine Linux's own advisory, published alongside the disclosure, gives specific guidance for anyone who thinks they may have already been hit: check /var/log/messages for sshd lines reading "Accepted password" for any account that has no password — whether that's alpine, root, or any account meant to be key-only. Alpine also recommends reviewing doas usage and any recently created users, files, or authorized keys on systems that were internet-exposed during the window before patching, and states plainly that if you cannot rule out a login during that window, the right posture is to treat the machine as compromised rather than assume it's clean.
Why this matters for anyone learning Linux administration
This is a genuinely good case study for exactly the kind of thinking Linux administration training is meant to build: knowing which authentication mechanism your system is actually using (PAM versus not), knowing which C library your distribution is built against and why that choice has security consequences you wouldn't expect, and understanding that a "disabled" account isn't automatically a safe account if the thing enforcing that disablement has a bug in it. It also reinforces a habit worth having regardless of this specific CVE: knowing exactly which of your systems — containers included — are running Alpine or another musl-based distribution, since that's the one fact that determines whether this bug applies to you at all. That depth of understanding — not memorizing CVE numbers, but actually knowing how authentication works under the hood so a bug like this one is recognizable the moment it's disclosed — is what Innovative Academy's Linux Administration course in Bangalore spends its time on. Contact for course details: 8447712333.
What we're not claiming
We're not claiming this affects glibc-based distributions, Dropbear, or any password-authenticated account with a genuine password hash — The Hacker News' reporting is explicit that none of those are in scope, and this piece isn't extending the bug's reach beyond what's documented. We're not claiming the exploitation attempts reported by Previdian were necessarily all against production systems rather than honeypots — the article's own wording doesn't rule out that some or all of the counted attempts targeted monitored decoy systems rather than real infrastructure, so treat the 20-attempts, 13-country figure as observed scanning activity, not confirmed real-world breaches. And we're not asserting the precise internal mechanics of musl's hash-computation failure beyond what's been publicly described — the reporting establishes that an oversized password causes the hash step to return an asterisk that then matches a disabled account's stored hash, but a full code-level explanation of musl's crypt() internals is a deeper technical question than this piece addresses; check OpenSSH's own 10.6p2 release notes or musl's own advisory if you need that level of detail.