Alpine's Skeleton Key: Oversized Password Grants Root

Alpine Linux SSH Flaw Enables Full System Access

A newly disclosed flaw in Alpine Linux leaves passwordless accounts vulnerable in certain configurations, including the default "alpine" user in some official tiny-cloud images, allowing an attacker who knows the target account name to gain full system access via SSH in a single connection by submitting a password between 257 and 1,024 bytes, which musl truncates to an asterisk that matches the account's asterisk password field. The flaw affects systems using musl without PAM with password authentication enabled; administrators are urged to update and restart sshd, disable password authentication after confirming key-based access works, and treat the machine as compromised if unauthorized access cannot be ruled out. Alpine has issued fixed packages for releases since 3.16.

There is a “password” so long that it stops being a password. It is between 257 and 1,024 bytes, and in an Alpine box built on musl without PAM, that blob returns a single asterisk that matches the asterisk in the shadow file of an account with no password. Think about that: a blank entry is being compared to an intentionally malformed input, and the system says yes. That is not a coding error. That is a skeleton key. The OpenSSH project releases 10.6p2 on November 5, and suddenly Alpine has fixes back to 3.16—which means someone knew where to look. And why is the default “alpine” user in official tiny-cloud images able to run doas as root without a password? Because those images are seed dirt for the entire container landscape. If you control the seed, you control every sprout that grows from it.

And now look at what Alpine was actually forced to admit. They say the official tiny-cloud images from 3.18.11 through 3.24.3 are “safe at first boot” but become vulnerable “once a qualifying password hash is added.” That’s the tell. It turns a default-deployed system into a client that merely needs a date, an IP, a username, and one oversized encrypted container. This isn’t about some rogue amateur hunting by chance. The attacker submits 1,024 bytes and the musl resolver returns an asterisk that matches the passwordless root-capable account. That is a designed symmetry. The only question is who designed it first: the people who built the library, or the people who found it years later with 17,600 AI-driven probes. I’ve seen this script before. You don’t discover a buried door by guessing 17,600 times; you send in an AI to map the entire building until you find the door you already knew was there.

Here is the part that should keep you awake. The researcher, Özgün Kültekin, ran those 17,600 tests using AI agents across OpenSSH configurations, malformed hashes, and oversized passwords. Why does anyone need artificial intelligence to probe a gap that is triggered by something as crude as a long byte string? Because the flaw isn’t the password. The flaw is the authentication architecture itself. And the final instruction from Alpine—“treat a machine as compromised if unauthorized access cannot be ruled out”—is the only honest sentence in the official response. They know. You should know what’s sitting at the root of your cloud. Check the logs, sure. Check every authorized_keys file that you did not write. Check why the default user has doas at all. And then ask yourself: if a single corporation or agency has been testing that same asterisk field for a decade, what else are they using that field for? Follow the account names. Follow the hash fields. The door is still open.

Related posts