Hi there, I’ve been reading up on selfhosting for a couple of weeks now and I got my feet wet with a couple of things.
However, before really getting serious with it, I feel I need to get down the basics and make sure that my server will not end up a security hazard. My final goal would be to self-host my socials (Mastodon, Lemmy, Matrix) - just for myself.
What basic security do I need to have in place, considering these services? I’ll be running this on a VPS and so far I consider the following: disable password login (login with ssh key only) then set up nginx, fail2ban, and a basic firewall. I’d try to close all ports that are not required for the services I run. I’ll also change ssh port from 22 to something else and close port 22 as well.
Would this be a sufficient basis, or am I missing something crucial?
Bonus question: do you know of good tutorials to learn the above stuff? I’ve been following the guides on DigitalOcean (e.g. https://www.digitalocean.com/community/tutorials/how-to-protect-an-nginx-server-with-fail2ban-on-ubuntu-20-04) and they seem decent enough - but I think I’ll need to get into more depth than that :)
Changing the port seems like a pointless step, just disallow access from everywhere and allow only from select IPs. Port scanners will scan all open ports and will detect that it’s ssh, regardless of port number.
Agreed. Security through obscurity is a fallacy.
If OP just wants to use it himself, a good idea might be to setup a VPN service and only allow the other services to be used from the VPN. That can be done with wire guard and a reverse proxy for example.
While I do completely agree, changing ports is more about getting rid of low-hanging fruit so some script kiddie doesn’t get into 22. But again I do agree with everything you said.
Just firewall the port and there’s no difference for your hypothetical script kiddies. Don’t ever do security by obscurity.
@animist @PlexSheep Agreed. Security through obscurity just delays a breach, for about 2 mins, while the port scan runs.
Of you have Skript kiddies logging in successfully on 22, you have way different problems.
Of course, changing a port number is not a good security improvement even in the realm of security through obscurity.
Changing the default port isn’t any more secure, but you’ll cut down on access attempts which will take some load off of your VM. It’s something that could be worth doing if you’re running a very cheap VPS.
It’s well worth it for this reason alone. Unless you’re running a honeypot, getting that noise out of your access logs can provide more actionable data to work with for more persistent threats.
You could try zerotier and only allow access from your ZT IPs. Thats how I set mine up, along with pointing cloudflare etc at it as well
Maybe try out FreedomBox?
freedomboxis a Debian package which automatically sets upapache2,firewalld,fail2banand Letʼs Encrypt. It also automatically adds pre-canned configuration files for applications you install with it (e.g. Mediawiki, WordPress, Matrix, Postfix/Dovecot). The theoretical goal of FreedomBox is to allow anyone to set up a webserver and administer it via a webUI. So, although I would say itʼs not quite there yet for command-line-illiterate users, I have found the software useful as a turnkey server to see what makes certain web applications tick, albeït in mostly vanilla form.For example, after installing a new app like WordPress, you could examine what exactly the FreedomBox scripts changed in the
/etc/apache2/or/etc/fail2ban/configuration files.Your plan is solid. The important thing is that you configure those things correctly, but you’re following guides so that should be ok. It’s on a VPS so there’s no threat to your home network, and none of those services pose a significant risk to you even if they were compromised so there’s no reason to go overboard.
If I had any further advice to give it’d be:
-
Change any default usernames and passwords that any of your apps/databases use.
-
Use randomly generated passwords for all service accounts. So that if you do find yourself compromised, they don’t then know a password that you’ve reused somewhere else (like your email account).
-
Run those services using something like Docker with no access to each other.
-
Utilize your VPS provider’s cloud firewall if they have one. If you’re paying for a cheap VM, it shouldn’t need to deal with all the general firewalling from the internet. VPS providers often have free cloud firewalls you can offload that work to.
Docker is the way to go. More often than not self-hosted stuff already has docker instructions, and by design it doesn’t mount your entire drive or give access to really anything on your system unless defined explicitly, even networks are isolated iirc. OP, get educated on what docker is and what flags it has so you can easily see what has access to what before even spinning something up.
-
Look into hardening measures implemented by https://github.com/dev-sec/ansible-collection-hardening/
I recommended reading through each role and decide which measures you want to implement (and port them to you own management system if you don’t use ansible)
In addition to what you mentioned, setup logcheck to email you unexpected logs. It does require a bit of time and fine tuning to make it ignore expected logs, but in terms of security measures it’s very powerful. I get an email every time I log in, incorrectly type my sudo password, etc. Sounds very verbose, but it also means it’s verbose when something unexpected is happening which is what you want security-wise. A nice side effect of having to craft the regexes of what logs to ignore is that I know better what’s running on my server :)
