On this page (8 sections)
A flaw in OpenSSH's sshd, tracked as CVE-2026-123158 and nicknamed "sshole", lets a remote attacker log in to certain passwordless accounts by sending a very long password. It affects OpenSSH before 10.6p2 on Linux systems that use the musl C library without PAM, with Alpine Linux as the best-known example. The fix shipped on November 5, 2026.
Threat intelligence firm Previdian reported exploitation less than nine hours after the CVE was published, according to Help Net Security. On November 9, 2026, the flaw was added to the CISA Known Exploited Vulnerabilities (KEV) catalog, as shown on the NVD entry. If you run Alpine-based servers, containers or appliances with SSH exposed, this is worth checking today.
What happened
The bug was found by researcher Özgün Kültekin, who published a write-up at sshhole.com. The OpenSSH fix landed in version 10.6p2, and the Alpine Linux advisory describes the impact bluntly: an attacker who can reach the SSH port can log in without knowing any password, in one connection, every time.
According to Help Net Security, Previdian founder Ryan Dewhurst said the first observed attempt arrived under nine hours after CVE publication. Previdian confirmed root access on a honeypot running real Alpine Linux and OpenSSH. The source address ran short reconnaissance commands (id, uname, hostname) over non-interactive SSH sessions. At that point Previdian had counted eight attempts, all from one Japan-based address and all seen by one sensor.
Exploitation is growing and CISA has listed it
The picture has widened since the first reports. The Hacker News updated its coverage on November 8 to say that Previdian's count had risen to 20 exploitation attempts from 13 unique IP addresses, located in Denmark, Iran, Japan, the Netherlands, Norway, Singapore, Sweden, Switzerland and the U.S. This is still reporting from a single vendor's telemetry.
A second, independent observer points the same way. The SANS Internet Storm Center reported on November 8 that its Cowrie honeypots, which record only the first 255 characters of SSH passwords, saw a rise in unusually long passwords. Daily counts went from 0 through October 31 to 232 on November 5 and 713 on November 6, then fell to 246 on November 7 and 51 on November 8. SANS says the first genuine sshole attempt arrived on November 5 using an all-"A" password, and that earlier long passwords on November 1 did not appear related to the exploit (they looked like buggy brute-force scripts pasting shell commands into the password field). SANS also said it expected quite a few of the attempts to have succeeded on its honeypots and planned to follow up with details.
On November 9, 2026, CISA added the flaw to its KEV catalog, listing it as an OpenBSD OpenSSH unexpected status code or return value vulnerability. CISA lists a due date of November 12, 2026 for federal agencies, and its record marks known ransomware use as "Unknown". Our earlier statement that CISA had not listed the flaw is no longer accurate.
Technical details in plain terms
Accounts that should never accept a password, such as root or a key-only deploy user, usually have a "*" in the password field of /etc/shadow instead of a hash. Two older design choices then collide, as the researcher explains it:
- Since 2016, to stop attackers from discovering valid usernames through timing differences (CVE-2016-6210), sshd hashes the supplied password anyway, borrowing another account's salt when the target account has no usable hash.
- musl's crypt() function has refused SHA-256 and SHA-512 passwords longer than 256 bytes since 2012, and on failure it returns "*".
- That failure value matches the "*" stored for the passwordless account, so sshd treats the password as correct.
The NVD description says that for some account configurations, any password between 256 and 1024 characters succeeds; The Hacker News words the range as 257 to 1,024 bytes. Either way, the trigger is an oversized password, far longer than normal. The weakness class is CWE-394, Unexpected Status Code or Return Value: a legitimate error result is not handled the way the calling code expects. The researcher reports testing many sshd build combinations, and only the musl build without PAM accepted logins it should have refused.
The attack does not bypass passwords on accounts that have valid password hashes. It needs at least one account with "*" as its password hash, and a successful login carries that account's existing permissions.
Who is affected
Exposure needs several conditions at once, per the reporting:
- OpenSSH before 10.6p2 on Linux with musl libc and no PAM. Alpine builds its default openssh-server package this way.
- Password authentication enabled, which is the OpenSSH default.
- An account with only "*" in its password field, plus another local account that has a password hash.
- Other musl-based systems worth checking, according to the researcher, include OpenWrt, Buildroot, Yocto and many Docker images.
Key-only users created without a password and directory users pulled in through musl-nscd fit the profile of the target account, according to SOCRadar. Root ships with "*" since about Alpine 3.16, but OpenSSH's default PermitRootLogin setting keeps root out of reach unless an admin changed it. The larger concern is indirect: on Alpine's official AWS and Azure images, the default alpine user has passwordless doas, so logging in as that user effectively means root. SOCRadar lists affected tiny-cloud images as 3.18.11 through 3.24.3.
Per the researcher, systems built on glibc (Debian, Ubuntu, Fedora), setups that authenticate through PAM, Dropbear and OpenBSD are not affected.
How to check your exposure
First, confirm whether the host is Alpine or another musl system. This read-only command prints the OS release information; look for Alpine or another musl-based distribution.
cat /etc/os-releaseNext, check the installed OpenSSH version. A version older than 10.6p2 on a musl system without PAM is in scope. Note that distributions may backport fixes, so also compare against the vendor advisory.
ssh -VOn Alpine, this lists the installed OpenSSH packages and their versions.
apk info -v | grep -i opensshFinally, look for signs of a break-in. Alpine and SOCRadar recommend searching the system log for successful password logins by accounts that have no password.
grep "Accepted password for" /var/log/messagesWhat to do now
- Identify every musl-based host, container image and appliance with SSH reachable (Alpine, OpenWrt, Buildroot, Yocto).
- Update to OpenSSH 10.6p2 or your distribution's patched package.
- Restart sshd after updating so the fixed binary is actually running.
- If you cannot patch immediately, set PasswordAuthentication to no, or disable SSH on these systems, as OpenSSH developer Damien Miller urged.
- Move to key-only logins and restrict SSH to trusted addresses with a firewall or security group.
- Review logs for "Accepted password for" entries naming passwordless accounts such as root, deploy or default cloud users, covering at least the days since November 1.
- Also review doas use and recently created users, files and authorized keys, as Alpine advises.
- If you cannot rule out a login, treat the machine as compromised and follow a recovery plan.
Updating on Alpine uses the standard package manager. This refreshes the package index and upgrades OpenSSH; afterwards, confirm the version shown by the check above.
apk update && apk upgrade openssh-serverThen restart the service so the patched binary takes over.
rc-service sshd restartAs a mitigation, set the following in /etc/ssh/sshd_config and restart sshd. It turns off password logins entirely, so make sure key-based access works first.
PasswordAuthentication noDetection and monitoring afterwards
Previdian published one observed source address, 89.116.89[.]18, but a single IP is a weak indicator since attackers change infrastructure, and the later reports describe 13 different addresses. SANS also listed the first scanning sources it saw, spread across many networks and countries. Better signals are successful password logins for accounts that should be key-only, very long password attempts if your logging captures them (many honeypots and logs truncate long passwords), and short bursts of id, uname or hostname commands from new sessions. If you do find an unexpected login, see our guide on what to do when a website is hacked.
Exposed SSH is also easy to forget. Webforta Monitor can watch uptime, certificate and DNS changes on your public hosts, and the Webforta Scan gives a read-only outside view of what your verified sites expose. Neither replaces patching the server itself.
Timeline
| Date | Event |
|---|---|
| 2026-11-01 | SANS honeypots begin seeing long passwords, which SANS says did not appear related to the exploit |
| 2026-11-05 | OpenSSH 10.6p2 released; CVE-2026-123158 published; Alpine advisory issued; first genuine sshole attempt seen by SANS |
| 2026-11-05 | Alpine reportedly pushed patched packages for branches 3.16 to edge (per SOCRadar) |
| 2026-11-06 | Previdian reports first attempt under nine hours after CVE publication and a successful honeypot login |
| 2026-11-08 | Previdian's count reported at 20 attempts from 13 IP addresses; SANS publishes honeypot data |
| 2026-11-09 | CVE-2026-123158 added to the CISA KEV catalog (due date 2026-11-12) |
Story updates
CVE-2026-123158 was added to the CISA KEV catalog on November 9, 2026, confirming active exploitation. The article now also includes Previdian's updated count (20 attempts from 13 IPs) and SANS ISC honeypot data, and corrects the earlier statement that it was not in KEV.
First published.
Sources
This article is based on the public sources below. Details can change as vendors and researchers publish more; check the original advisories before acting on version numbers.
- CVE-2026-123158 - GitHub Advisory Database
- OpenBSD OpenSSH Unexpected Status Code or Return Value Vulnerability (CVE-2026-123158) is being actively exploited
- Scans for Sshole Vulnerability (CVE-2026-123158) - SANS ISC
- Hackers Target OpenSSH SSHole Flaw Within Hours of Disclosure, Gain Root on Honeypot
- OpenSSH flaw exploited in the wild just hours after disclosure (CVE-2026-123158) - Help Net Security
