tldr: I’d like to set up a reverse proxy with a domain and an SSL cert so my partner and I can access a few selfhosted services on the internet but I’m not sure what the best/safest way to do it is. Asking my partner to use tailsclae or wireguard is asking too much unfortunately. I was curious to know what you all recommend.
I have some services running on my LAN that I currently access via tailscale. Some of these services would see some benefit from being accessible on the internet (ex. Immich sharing via a link, switching over from Plex to Jellyfin without requiring my family to learn how to use a VPN, homeassistant voice stuff, etc.) but I’m kind of unsure what the best approach is. Hosting services on the internet has risk and I’d like to reduce that risk as much as possible.
-
I know a reverse proxy would be beneficial here so I can put all the services on one box and access them via subdomains but where should I host that proxy? On my LAN using a dynamic DNS service? In the cloud? If in the cloud, should I avoid a plan where you share cpu resources with other users and get a dedicated box?
-
Should I purchase a memorable domain or a domain with a random string of characters so no one could reasonably guess it? Does it matter?
-
What’s the best way to geo-restrict access? Fail2ban? Realistically, the only people that I might give access to live within a couple hundred miles of me.
-
Any other tips or info you care to share would be greatly appreciated.
-
Feel free to talk me out of it as well.
I use traefik with a wildcard domain pointing to a Tailscale IP for services I don’t want to be public. For the services I want to be publicly available I use cloudflare tunnels.
if you know/use docker, the solution that has been the most straightforward for me is SWAG. the setup process is fairly easy when combined with registering your domain with Porkbun, as they allow free API access needed for obtaining top-level (
example.com) as well as wildcard (*.example.com) SSL certificates.along with that, exposing a new service is fairly easy with the plethora of already included nginx configs for services like Nextcloud, Syncthing, etc.
Why do so many people do this incorrectly. Unless you are actually serving a public then you don’t need to open anything other than a WireGuard tunnel. My phone automatically connects to WireGuard as soon as I disconnect from my home WiFi so I have access to every single one of my services and only have to expose one port and service.
If you are going through setting up caddy or nginx proxy manager or anything else and you’re not serving a public… you’re dumb.
What are you using to auto connect to VPN when you disconnect from your home wifi?
The Wireguard iOS app has an “on-demand” toggle that automatically connects when certain conditions are met (on cellular, on wifi, exclude certain networks, etc)
WG Tunnel does that natively, you can whitelist some wifis and auto connect on other and optionally on mobile data
Tasker on android, bit faffy and shouldn’t at all be necisary
Why is it too much asking your partner to use wireguard? I installed wireguard for my wife on her iPhone, she can access everything in our home network like she was at home, and she doesn’t even know that she is using VPN.
A few reasons
- My partner has plenty of hobbies but sys-admin isn’t one of them. I know I’ll show them how to turn off wireguard to troubleshoot why “the internet isn’t working” but eventually they would forget. Shit happens, sometimes servers go down and sometimes turning off wireguard would allow the internet to work lol
- I’m a worrier. If there was an emergency, my partner needed to access the internet but couldn’t because my DNS server went down, my wireguard server went down, my ISP shit the bed, our home power went out, etc., and they forgot about the VPN, I’d feel terrible.
- I was a little too ambitious when I first got into self hosting. I set up services and shared them before I was ready and ended up resetting them constantly for various reasons. For example, my Plex server is on it’s 12th iteration. My partner is understandably weary to try stuff I’ve set up. I’m at a point where I don’t introduce them to a service I set up unless accessing it is no different than using an app (like the Homeassistant app) or visiting a website. That intermediary step of ensuring the VPN is on and functional before accessing the service is more than I’d prefer to ask of them
Telling my partner to visit a website seems easy, they visit websites every day, but they don’t use a VPN everyday and they don’t care to.
you’re talking to a community of admins that force their family to “use the thing”. they can’t understand why anyone can’t debug tech issues because they have surrounded themselves with people who can.
I get it, my wife isn’t technical at all. she gets online about once a week to check email. I couldn’t even begin to explain to her how to debug her connection problems past turn it off and on again.
so, to simplify things, she doesn’t connect to the home network outside of the home network. but I was able to teach her how to download movies/shows from Plex to her phone and I was able to explain why ads show up on her apps when she’s out of the house.
it’s not perfect, but it’s the best I can give her with her understanding of the technology. knowing the limitations of your user base is just as important as developing the tools they will use and how they will access them.
I get where the original commenter is coming from. A VPN is easy to use, why not have my partner just use the VPN? But like, try adding something to your routine that you don’t care about or aren’t interested in. It’s an uphill battle and not every hill is worth dying on.
All that to say, I appreciate your comment.
- I don’t think this is a problem with tailscale but you should check. Also you don’t have to pipe all the traffic through your tunnel. In the allowed IPs you can specify only your subnet so that everything else leaves via the default gateway.
- in the DNS server field in your WireGuard config you can specify anything, doesn’t have to be RFC1918 compliant. 1.1.1.1 will work too
- At the end of the day, a threat model is always gonna be security vs. convenience. Plex was used as an attack vector in the past as most most people don’t rush to patch it (and rightfully so, there are countless horror stories of PMS updates breaking the whole thing entirely). If you trust that you know what you’re doing, and trust the applications you’re running to treat security seriously (hint: Plex doesn’t) then go ahead, set up your reverse proxy server of choice (easiest would be Traefik, but if you need more robustness then nginx is still king) and open 443 to the internet.
For point number 2, security through obscurity is not security.
Besides, all issued certificates are logged publicly. You can search them here https://crt.shNginx Proxy Manager is easy to set up and will do LE acme certs, has a nice GUI to manage it.
If it’s just access to your stuff for people you trust, use tailscale or wireguard (or some other VPN of your choice) instead of opening ports to the wild internet.
Much less riskI came here to upvote the post that mentions haproxy, but I can’t see it, so I’m resorting to writing one!
Haproxy is super fast, highly configurable, and if you don’t have the config nailed down just right won’t start so you know you’ve messed something up right away :-)
It will handle encryption too, so you don’t need to bother changing the config on your internal server, just tweak your firewall rules to let whatever box you have haproxy running on (you have a DMZ, right?) see the server, and you are good to go.
Google and stackexchange are your friends for config snippets. And I find the actual documentation is good too.
Configure it with certificates from let’s encrypt and you are off to the races.
Tailscale is very popular among people I know who have similar problems. Supposedly it’s pretty transparent and easy to use.
If you want to do it yourself, setting up dyndns and a wireguard node on your network (with the wireguard udp port forwarded to it) is probably the easiest path. The official wireguard vpn app is pretty good at least for android and mac, and for a linux client you can just set up the wireguard thing directly. There are pretty good tutorials for this iirc.
Some dns name pointing to your home IP might in theory be an indication to potential hackers that there’s something there, but just having an alive IP on the internet will already get you malicious scans. Wireguard doesn’t respond unless the incoming packet is properly signed so it doesn’t show up in a regular scan.
Geo-restriction might just give a false sense of security. Fail2ban is probably overkill for a single udp port. Better to invest in having automatic security upgrades on and making your internal network more zero trust
If security is one of your concerns, search for “HTTP client side certificates”. TL;DR: you can create certificates to authenticate the client and configure the server to allow connections only from trusted devices. It adds extra security because attackers cannot leverage known vulnerabilities on the services you host since they are blocked at http level.
It is a little difficult to find good and updated documentation but I managed to make it work with nginx. The downside is that Firefox mobile doesn’t support them, but Firefox PC and Chrome have no issues.
Of course you want also a server side certificate, the easiest way is to get it from Let’s Encrypt
That’s interesting, I didn’t know that was a thing. I’ll look into it, thanks!
I remember that I started by following these two guides.
https://fardog.io/blog/2017/12/30/client-side-certificate-authentication-with-nginx/
https://stackoverflow.com/questions/7768593/
something I’m not sure it is mentioned here is that android (at lest the version on my phone) accepts only a legacy format for certificates and the error message when you try to import the new format is totally opaque. If you cannot import it there just check openssl flags to change the export format.
Caddy with cloudflare support in a docker container.
Does Caddy have an OWASP plugin like nginx?
I don’t use it, but it looks like yes.
This the solution.
Caddy is simple.
I currently have a nginx docker container and certbot docker container that I have working but don’t have in production. No extra features, just a barebones reverse proxy with an ssl cert. Knowing that, I read through Caddy’s homepage but since I’ve never put an internet facing service into production, it’s not obvious to me what features I need or what I’m missing out on. Do you mind sharing what the quality of life improvements you benefit from with Caddy are?
What caddy does are automatic certs. You set up your web-portal and make a wildcard subdoman that points to your portal. Then you just enter two lines in the config and your new app is up. Lets say you want to put your hone assistant there. You could add hass.portal.domain.tld {reverse_proxy internal.ip:8123 } and it works. Possible with other setups too, but its no hassle
Honestly, if you know nginx just stick with it. There’s nothing to be gained by learning a new proxy.
Use Mozilla’s SSL generator if you want to harden nginx (or any proxy you choose)- https://ssl-config.mozilla.org/
I didn’t know about that tool. Thanks for sharing
I never went too far down the nginx route, so I can’t really compare the two. I ended up with caddy because I self-host vaultwarden and it really doesn’t like running over http (for obvious reasons) and caddy was the instruction set I found and understood first.
I don’t make a lot of what I host available to the wider internet, for the ones that I do, I recently migrated to using a Cloudflare tunnel to deal with the internet at large, but still have it come through caddy once it hits my server to get ssl. For everything else I have a headscale server in Oracle’s free tier that all my internal services connect to.
Either tailscale or cloudflare tunnels are the most adapted solution as other comments said.
For tailscale, as you already set it up, just make sure you have an exit node where your services are. I had to do a bit of tinkering to make sure that the ips were resolved : its just an argument to the tailscale command.
But if you dont want to use tailscale because its to complicated to your partner, then cloudlfare tunnels is the other way to go.
How it works is by creating a tunnel between your services and cloudlare, kind of how a vpn would work. You usually use the cloudlfared CLI or directly throught Cloudflare’s website to configure the tunnel. You should have a DNS imported to cloudflare by the way, because you have to do a binding such as : service.mydns.com -> myservice.local Cloudlfare can resolve your local service and expose it to a public url.
Just so you know, cloudlfare tunnels are free for some of that usage, however cloudlfare has the keys for your ssl traffic, so they in theory could have a look at your requests.
best of luck for the setup !
Thanks for the info, I appreciate it
How do you handle SSL certs and internet access in your setup?
I have NPM running as “gateway” between my LAN and the Internet and let handle it all of my vertificates using the built-in Let’s Encrypt features. None of my hosted applications know anything about certificates in their Docker containers.
As for your questions:
- You can and should – it makes managing the applications much easier. You should use some containerization. Subdomains and correct routing will be done by the reverse proxy. You basically tell the proxy “when a request for foo.example.com comes in, forward it to myserver.local, port 12345” where 12345 is the port the container communicates over.
- 100% depends on your use case. I purchased a domain because I host stuff for external access, too. I just have my setup to report it’s external IP address to my domain provider. It basically is some dynamic DNS service but with a “real domain”. If you plan to just host for yourself and your friends, some generic subdomain from a dynamic DNS service would do the trick. (Using NPMs Let’s Encrypt configuration will work with that, too.)
- You can’t. Every georestricting can be circumvented. If you want to restrict access, use HTTP basic auth. You can set that up using NPM, too. So users authenticate against NPM and only when it was successful,m the routing to the actual content will be done.
- You might want to look into Cloudflare Tunnel to hide your real IP address and protect against DDoS attacks.
- No 🙂
“NPM” node package manager?
- Yeah I’ve been playing around with docker and a domain to see how all that worked. Got the subdomains to work and everything, just don’t have them pointing to services yet.
- I’m definitely interested in the authentication part here. Do you have an tutorials you could share?
- Will do, thanks
- ❤️
I don’t know how markdown works. that should be 1,3,4,5
Authentication with NPM is pretty straightforward. You basically just configure an ACL, add your users, and configure the proxy host to use that ACL.
I found this video explaining it: https://youtu.be/0CSvMUJEXIw?t=62
NPM unfortunately has a long-term bug since 2020, that needs you to add a specific configuration when setting up the ACL as shown in the video.
At the point where he is on the “Access” tab with all the allow and deny entries, you need to add an allow entry with
0.0.0.0/0as IP address.
Other than that, the setup shown in the video works in the most recent version.
nginx proxy manager
there’s so many acronyms. Thanks
I was reading this and thinking node package manager too and I was both confused and concerned that somebody would sit all of their security on node package manager!
That makes much more sense 🙂
or a domain with a random string of characters so no one could reasonably guess it? Does it matter?
That does not work. As soon as you get SSL certificates, expect the domain name to be public knowledge, especially with Let’s Encrypt and all other certificate authorities with transparency logs. As a general rule, don’t rely on something to be hidden from others as a security measure.
It is possible to get wildcard certificates from LetsEnrcypt which doesn’t give anyone information on which subdomains are valid as your reverse proxy would handle that. Still arguably security through obscurity, but it does make it substantially harder for anyone who can’t intercept traffic between the client and server.
Damn, I didn’t realize they had public logs like that. Thanks for the heads up
Https://crt.sh would make anyone who thought obscurity would be a solution poop themselves.
The biggest reason to use VPN is that some ISPs may take issue with you running a web server over a residential service when they see incoming HTTP requests to your IP. If you don’t want to require VPN, then Cloudflare tunnels are perfect for this and they also solve the need for dynamic DNS if you want to use static domain because your domain points to the Cloudflare edge servers and they route it to you wherever your tunnel endpoint is running.
Past that, Traefik is a great reverse proxy that can manage getting LetsEnrcypt SSL certificates for you even with wildcard domains and would still work fine with dynamic DNS.
ISPs shouldn’t care unless it is explicitly prohibited in the contract. (I’ve never seen this)
I still wouldn’t expose anything locally though since you would need to pay for a static IP.
Instead, I just use a VPS with Wireguard and a reverse proxy.
Do you mind giving a high level overview of what a Cloudlfare tunnel is doing? Like, what’s connected to what and how does the data flow? I’ve seen cloudflare mentioned a few other times in the comments here. I know Cloudflare offers DNS services via their 1.1.1.1 and 1.0.0.1 IPs and I also know they somehow offer DDoS protection (although I’m not sure how exactly. caching?). However, that’s the limit of my knowledge of Cloudflare
Basically the Cloudflare tunnel client connects from the computer running your services (or proxy) out to Cloudflare’s edge servers and your DNS hostname is set to the IP of one of Cloudflare’s edge servers. Cloudflare acts like a reverse proxy by sending incoming SSL requests for your hostname to your tunnel client through their own network. The DNS record doesn’t expose your public IP and the Cloudflare tunnel client easily works behind firewalls, NAT, and doesn’t need a static IP because it connects outbound to Cloudflare’s network.
The biggest limitation is that this only works for SSL traffic because it can be routed by hostname in the SNI without needing a client on the client side. They do offer tunnels for other connections, but that requires their client running on both sides so it’s more like a traditional VPN again.
I use a central nginx container to redirect to all my other services using a wildcard let’s encrypt cert for my internal domain from acme.sh and I access it all externally using a tailscale exit node. The only publicly accessible service that I run is my Lemmy instance. That uses a cloudflare tunnel and is isolated in it’s own vlan.
TBH I’m still not really happy having any externally accessible service at all. I know enough about security to know that I don’t know enough to secure against much anything. I’ve been thinking about moving the Lemmy instance to a vps so it can be someone else’s problem if something bad leaks out.
Don’t fret, not even Microsoft does.
You’re not as valuable as a target as Microsoft.
It’s just about risk tokerance. The only way to avoid risk is to not play the game.
wildcard let’s encrypt cert
I know what “wildcard” and “let’s encrypt cert” are separately but not together. What’s going on with that?
How do you have your tailscale stuff working with ssl? And why did you set up ssl if you were accessing via tailscale anyway? I’m not grilling you here, just interested.
I know enough about security to know that I don’t know enough to secure against much anything
I feel that. I keep meaning to set up something like nagios for monitoring and just haven’t gotten around to it yet.
So when I ask Let’s Encrypt for a cert, I ask for *.int.teuto.icu instead of specifically jellyfin.int.teuto.icu, that way I can use the same cert for any internally running service. Mostly I use SSL on everything to make browsers complain less. There isn’t much security benefit on a local network. I suppose it makes harder to spoof on an external network, but I don’t think that’s a serious threat for a home net. I used to use home.lan for all of my services, but that has the drawback of redirecting to a search by default on most browsers. I have my tailscale exit node running on my router and it just works with SSL like anything else.
Ok so I currently have a cert set up to work with:
www.domain.com (some browsers seemingly didn’t like it if I didn’t have www)
Are you saying I could just configure it like this:
*.domain.com
The idea of not having to keep updating the cert with new subdomains (and potentially break something in the process) is really appealing
Yes. If you’re using lets encrypt then note that they do not support wildcard certs with the HTTP-01 challenge type. You will need to use the DNS-01 challenge type. To utilize it you would need a domain registrar that supports api dns updates like cloudflare and then you can use the acme.sh package. Here is an example guide i found.
Note that you could still request multiple explicit subdomains in the same issue/renew commands so it’s not a huge deal either way but the wildcard will be more seamless in the future if you don’t know what other services you might want to selfhost.
awesome, thanks for the info
nixos with nginx services does all proxying and ssl stuff, fail2ban is there as well
I know I should learn NixOS, I even tried for a few hours one evening but god damn, the barrier to entry is just a little too high for me at the moment 🫤
i guess you were able to install the os ok? are you using proxmox or regular servers?
i can post an example configuration.nix for the proxy and container servers that might help. i have to admit debugging issues with configurations can be very tricky.
in terms of security i was always worried about getting hacked. the only protection for that was to make regular backups of data and config so i can restore services, and to create a dmz behind my isp router with a vlan switch and a small router just for my services to protect the rest of my home network
i guess you were able to install the os ok? are you using proxmox or regular servers?
I was. It was learning the Nix way of doing things that was just taking more time than i had anticipated. I’ll get around to it eventually though
I tried out proxmox years ago but besides the web interface, I didn’t understand why I should use it over Debian or Ubuntu. At the moment, I’m just using Ubuntu and docker containers. In previous setups, I was using KVMs too.
Correct me if I’m wrong, but don’t you have to reboot every time you change your Nix config? That was what was painful. Once it’s set up the way you want, it seemed great but getting to that point for a beginner was what put me off.
I would be interested to see the config though
