Adding a user to the docker group is equivalent with granting it passwordless root privileges, never do that unless you know precisely why and understand the risks involved.
On the positive side, forbidding password-based auth, installing and using a firewall and locking down the root account is good advice
Getting whatever you can (likely most, or all) into rootless podman, preferably with SELinux is worthwhile. Docker is a security nightmare.
I have a whole bunch of docker compose declared containers running on my server how much of an investment do you think it’d be to podman? I don’t like how insecure docker is but it seems like a pain to move.
Not going to lie, there is pain, but maintenance is much nicer. I’ve gotten to the point where I can just convert docker-compose to quadlet by eye, the syntax is familiar if you know systemd.
https://github.com/containers/podlet can be pointed at a docker-compose and spits out quadlets. It’s been a while but last I used it, it would get it right 80%+ of the time on complex things and get you 85%+ there when it failed. There’s also podman-compose which is a crutch but can establish compatibility.
/usr/libexec/podman/quadlet -dryrun -useris a useful diagnostic when conversions break.Honestly the issue is docker was never designed to be a security boundary. Docker is always from the beginning a way to deploy application without conflicting shared library requirements. Looser than isolation of VM, but also not as hard as every other existing solution at that time (dedicated chroot, static linking, appimage, flatpak)
Bookmarked 🥰
I would add changing the SSH port to one that isn’t 22. This isn’t any more secure, but does tamp down on automated scanner traffic.





