Back to blog
security

If RunMist Were Breached, What Could the Attacker Do to Your Server?

By Daniel

If RunMist Were Breached, What Could the Attacker Do to Your Server?

Every tool that deploys to your server has to get into it somehow. That makes the tool itself the most valuable target: break into the tool, and you have a door into every server it manages.

It is the first question worth asking any server panel, and most do not answer it on their front page. This is our answer.

The usual model

The common design is simple. When the panel sets up your server, it installs a key of its own and keeps it. From then on it can log in whenever it likes, often as root, for as long as the server exists.

It works, and it is easy to build. It also means a copy of the panel's database is a copy of the keys to every customer's server.

What we do instead

RunMist stores no key that opens your server. There is no SSH private key in our database, on our servers, or on yours. When a deploy or a maintenance job runs, it creates a throwaway key and gets a certificate for it that expires in 5 minutes. When the job ends, both are gone. A copy of our data opens nothing.

Our login is not root. The account RunMist uses can run a fixed set of maintenance commands, each one checked. It cannot open a root shell.

RunMist cannot give itself a new login. Your server decides who gets in, not our platform. It accepts one admin login from you, when the server is created or during its first day. After that, the only way to add a login is through your cloud provider's own console, which we do not control.

Your secrets are sealed to your server. Environment variables are encrypted so that only your server can open them. Our database holds nothing that decrypts them.

Apps are walled off from each other. Each project runs, installs its packages and builds as its own user. If one app is broken into through a bad dependency, the attacker reaches that app, not the others on the server.

Your own way in

None of this locks you out. It is your server, in your cloud account.

You never need to SSH in for daily work: deploys, logs, restarts, rollbacks, backups and domains are all in the dashboard. But if you want into the box, your admin login has full root, with a key only you hold. Think of it as the break-glass path. We never see that key.

What this does not cover

We would rather you hear the limits from us.

  • A live break-in is not the same as a stolen copy. Someone controlling our running platform could start the jobs the platform is allowed to start, for as long as they stay in. They would still not have root, and could not add a login. But they could deploy.
  • Apps on one server share its kernel. That is true of containers too. Security updates install every night, and a kernel fix restarts the server in a window you choose.
  • Your cloud account is yours to protect. Anyone with your provider login owns the server, whatever we do.
  • We are new. The design above is checked on real servers every week by automated runs, but it has not yet had an outside audit. When it has, we will publish the result.

Why we built it this way

Because you own the server. A tool you pay to deploy your apps should not also be a standing key to everything on it. If you stop using RunMist, the server keeps running, and nothing of ours is left holding a way in.