
Yesterday's post was about the flaw. Today's is about what happened next, because the gap between the two was nine hours.
On Thursday, November 5, OpenSSH shipped 10.6p2 to close SSHole, a password bypass in sshd on musl-based Linux (CVE-2026-123158, CVSS 8.1). At that point nobody had seen it used in the wild. The first exploitation attempt arrived less than nine hours after the CVE was published. By November 8, one threat intel firm had counted 20 attempts from 13 IP addresses across 9 countries, and confirmed root access on an Alpine honeypot.
The question now is whether the patch is actually running on your boxes, and whether anyone already walked in while you were deciding.
What changed since Thursday
Three things, all reported over the weekend:
- Previdian's sensors saw the first attempt under nine hours after the CVE went public. The earliest observed source, an IP in Japan, used oversized passwords to get in and ran
id,unameandhostnamethrough non-interactive SSH sessions. - By November 8 the count was 20 attempts from 13 unique IPs across nine countries, and Previdian confirmed root access on an Alpine honeypot.
- WatchTowr, which had predicted scanning by the weekend, says only botnets and crypto miners have picked it up so far. GreyNoise and Defused are reporting attempts too.
"Only botnets and miners so far" is not reassurance. It is the first wave, the automated one, the one that tries everything cheap.
The flaw in thirty seconds
In short: on musl-based Linux running sshd without PAM, which is how Alpine builds its default openssh-server package, a password longer than 256 bytes makes musl's crypt() fail and return an asterisk. An asterisk is also exactly what sits in the password field of every account with no password set. sshd compares the two, they match, and the login succeeds.
The attacker needs a username for an account whose password field is *: root on Alpine, a key-only user, or the alpine user in the official tiny-cloud images. The machine also needs at least one other account with a $5$ or $6$ hash, because sshd borrows that account's salt to process the attempt. One user with a password is enough to make every passwordless account reachable. The password has to be 257 to 1,024 bytes long, and it takes one connection.
Debian, Ubuntu, Fedora and Red Hat are not affected, because they use glibc with PAM. Neither is Dropbear, nor accounts locked with ! in /etc/shadow, nor most Alpine containers, which do not run sshd. The exposure is Alpine hosts and VMs that run sshd with password authentication on, which is the default.
How to tell if it knocked on your box
Patching is step one. Before that, or right after, check whether anyone already got in. Alpine's advisory gives the check, and it is short.
1. Confirm you are in scope. On each Alpine host, check the installed package against the fixed versions, 10.3_p1-r2 on 3.24 down to 9.9_p2-r1 on 3.21, with backports to 3.16. Then look at the accounts:
apk info openssh-server
grep ':\*:' /etc/shadow
grep -E '\$[56]\$' /etc/shadow
sshd -T | grep -i passwordauthentication
The first tells you what you are running. The second lists accounts with no password. The third shows whether any account carries a $5$ or $6$ hash, the salt the attack borrows. The fourth shows whether password authentication is on. If the middle two return lines and the fourth says yes, the box was reachable.
2. Read the logs. Alpine's advice: look in /var/log/messages for sshd lines reading "Accepted password for" an account that has no password. A successful password login for root, alpine, or any key-only user is not a normal event on a box like this. It is the signature.
grep 'Accepted password' /var/log/messages
Check every line against the accounts from step one. One hit for an account that should never accept a password is enough to act on.
3. Check for the aftermath. The observed attacks ran recon first, id, uname, hostname, then whatever came next. Review doas use, recently created users, files and authorized keys on any exposed machine, and look for unfamiliar processes, especially anything burning CPU the way a miner would. Check ~/.ssh/authorized_keys for every account, not just root.
4. If you cannot rule out a login, treat the machine as compromised. That is Alpine's line, not mine, and it is the right one. A password login that should have been impossible leaves no reliable way to prove nothing else happened. Rebuild or restore from a known-good image, rotate the keys, and do not spend the afternoon trying to prove a negative.
The fix, in order
If the logs are clean, close the hole properly. The order matters, because the step people skip is the one that matters most.
- Update the package:
apk upgrade -U -a. Fixed packages shipped November 5 for every branch from 3.16 to edge, including branches already out of support. Point releases 3.21.10, 3.22.8, 3.23.8 and 3.24.4 carry the fix. - Restart sshd:
rc-service sshd restart. Upgrading the package does not replace the running daemon. The old sshd keeps serving the old code until it restarts. This is the step that turns a patched package into a patched service, and the one most likely to be forgotten on a host you updated from a script. - Verify the running daemon is the new one: confirm the installed version with
apk info openssh-serverand the effective config withsshd -T | grep -i passwordauthentication.
If you cannot update yet, disable password logins instead: set PasswordAuthentication no in sshd_config and restart sshd. Confirm key login works in a second session before closing the first. Alpine's advisory has a drop-in file method for 3.19 and later.
One more check: the official tiny-cloud images from 3.18.11 through 3.24.3 ship an alpine user with a * password field and passwordless doas. They are safe on first boot and become exploitable the moment any $6$ hash exists, for example when an admin sets a console password. If you run those images, update or disable password auth on every running instance, not just the template.
The durable point
The interesting number in this story is not 20 attempts. It is nine hours. Nine hours from a CVE being published to the first exploitation attempt, on a flaw that had sat in the code since a 2016 fix for a different SSH bug. OpenSSH said it outright in the 10.6p2 release notes: bugs found by AI tools have been independently rediscovered by other researchers, which means adversaries who never report bugs can find them too. That is why the project is releasing more often instead of batching fixes.
The window between "patch available" and "exploited in the wild" is now measured in hours. A patch you have not installed is not a control. A patch you installed but did not restart into is not a control either. And a flaw that is not in CISA's KEV catalog yet, this one had not appeared there as of November 7, is still a flaw attackers are using.
Patch it, restart the daemon, read the logs. In that order, today.
Sources
- Alpine Linux: Security advisory: openssh-server password authentication bypass — November 5, 2026
- OpenSSH 10.6p2 release notes — November 5, 2026
- The Hacker News: Hackers Target OpenSSH SSHole Flaw Within Hours of Disclosure, Gain Root on Honeypot — November 6, 2026 (updated November 8, 2026)
- SecurityWeek: Hackers Exploiting 'SSHole' Authentication Bypass in OpenSSH — November 9, 2026
- Cybernoz: OpenSSH Rushes Patches for 'SSHole' Authentication Bypass — November 8, 2026



