Hello,
Some time ago, I started self-hosting applications, but only on my local network. So far, it’s working fine, but I can’t access them as soon as I go outside (which is completely normal).
For the past few days I’ve been looking for a relatively secure way of accessing my applications from outside.
I don’t need anyone but myself to have access to my applications, so from what I’ve understood, it’s not necessarily useful to set up a reverse-proxy in that case and it would be simpler to set up a VPN.
From what I’ve seen, Wireguard seems to be a good option. At first glance, I’d have to install it on the machine containing my applications, port-forward the Wireguard listening port and configure my other devices to access this machine through Wireguard
However, I don’t have enough hindsight to know whether this is a sufficient layer of security to at least prevent bots from accessing my data or compromising my machine.
I’ve also seen Wireguard-based solutions like Tailscale or Netbird that seem to make configuration easier, but I have a hard time knowing if it would really be useful in my case (and I don’t really get what else they are doing despite simplifying the setup).
Do you have any opinions on this? Are there any obvious security holes in what I’ve said? Is setting up a VPN really the solution in my case?
Thanks in advance for your answers!
I use a mixture of tailscale and zerotier. Both are pretty powerful.
Tailscale. You can make a free account and they have clients for most things. If you want to self host, Headscale.
If your traffic is pretty low, rent a VPS for $5/month or whatever and set up a Wireguard server on it, have your devices maintain a connection to it (search keepalive for Wireguard), and set up HAProxy to do SNI-based routing for your various subdomains to the appropriate device.
Benefits:
- you control everything, so switching to a new provider is as simple as copying configs instead of reconfiguring everything
- most VPN companies only route traffic going out, not in; you can probably find one that does, but it probably costs more than the DIY option
- easy to share with others, just give a URL
Downsides:
- more complicated to configure
- bandwidth limitations
If you only need access on devices you control, something like Tailscale could work.
Benefits:
- very simple setup - Tailscale supports a ton of things
- potentially free, depending on your needs
Downsides:
- no public access, so you’d need to configure every device that wants to access it
- you don’t control it, so if Tailscale goes evil, you’d need to change everything
I did the first and it works well.
I would not recommend relying on Tailscale. They have been soliciting a lot of venture capital lately and are probably going to go for an IPO sooner or later. I would not put a lot of trust in that company. The investors are going to want their money.
Not running anything myself but am part of a self hosting discord that swears by Netbird because its basically Tailscale but with a bunch more ease of use features apparently
I used wireguard self hosted for a bit but my work network is pretty locked down and I couldn’t find a UDP port that wasn’t blocked. How are you guys setting up wireguard in your home network? Or is it better to host it on a cloud VM?
I’m using tailscale right now because it punches through every firewall but I don’t like using external providers and I’m worried it will eventually enshittify. I have a cloudflare domain but I can’t really use any UDP port for my VPN as it’s blocked.
I got a VPS because I’m behind CGNAT, and then configured WireGuard on the server and HAProxy to proxy requests to my devices.
It works well for me.
Wireguard with WG Tunnel on my phone so it automatically connects when I leave my WiFi. Some Apps excluded to use it like Android Auto because it doesn’t work with an active vpn.
There’s no magic bullet here. If you want good defense against bots you should use fail2ban and/or crowdsec. Geoblocking is also worth looking into. You will always have to open a port if you are selfhosting a VPN and will need to take aforementioned steps (or alternatives) to secure it. I believe Tailscale is a very good alternative for people who don’t have time to do this as it does not (to the best of my knowledge) require you to expose a port.
I use Netbird (open source networking software from a German company) as it integrates well with Authentik and allows me to use the same SSO for VPN and most of my other services. Setting it up with Authentik and Nginx is a bit complicated but very well documented in my opinion. I do not have a positive experience of the official Android client but Jetbird is a nice alternative. Setting up DNS servers and network routes through peers is quite easy. Enrollment is also a breeze due to the Authentik integration.
Netbird is very nice and easy to use. Only downside is that the iOS app drains battery like crazy :(
WireGuard is the fastest method, it’s free, there is no reason not to use it.
I rarely if ever see ZeroTier mentioned as a solution, but it’s a self-hostable encrypted virtual mesh network (with a small free tier for corp-hosted), super secure, and really easy to setup. I use ZTnet instead of the free-tier corp-hosted controller
In that case, wireguard. I only occasionally need to access a service that’s not exposed to the internet, so I use
ssh -L, but that would be quite inconvenient for your own use case.I know tailscale exists but I’ve never used it, only tried wireguard on its own. Maybe there’s some huge benefit to using it but wireguard worked fine for me.
Run WireGuard on some home machine. (Does not need to be the machine the app you want to access is hosted on.)
Run WireGuard on your road warrior system.
There is no step 3.
I’m doing this right now from halfway around the world from my house and it’s been great. Been using iPhone, iPad, and macOS clients connected to linuxserver/WireGuard docker container.
Doesn’t that need like a static IP address, port forwarding and dealing all kind of network annoyances?
Recommending wireguard to people feels like recommending Arch to first time Linux users.
You don’t need a static IP address, but you do need a public IP address. You can use dynamic DNS to avoid having to keep track of your IP address. FreeDNS will work fine for a basic setup.
Wireguard is one of the easiest VPN servers to use. If you’re not using your ISP’s router, it may even have Wireguard built in.
Is wire guard a service you pay for? Otherwise how does wire guard in your home machine not need your router to forward ports to it? And then the remote client need to be pointed at your home’s external IP?
WireGuard is free. Obviously my instructions didn’t go into detail about specifically how to set everything up. Port forwarding is required. Knowing your servers external IP address is required. You also need electricity, an ISP subscription, a home server (preferably running Linux), so on and so forth. This is /c/selfhosted after all.
Yeah that’s fine. The steps were so simple I figured they could work without router config changes if they made some kind of connection handshake in a third party service’s server.
But given all that, I wonder if it makes sense to look into if your router has its own vpn server (or flash the firmware with one that does.)
Some routers even run WireGuard natively :) like for instance Ubiquiti. Personally I’d rather run it on my own server though because ubiquiti doesn’t have easy IAC features.
Apologies for the dumb noob question, but if your iOS device is VPNed to your home server, how does it access the open internet? Does it do this via the VPN to the home server?
Depends on the client configuration. If you route all the traffic through vpn (so, simplyfied, 0.0.0.0/0) then all their client device network traffic would go through their vpn server at home and is seen as coming from there; otherwise, if you only route specific addressess (like your home network private addressess only) then only those go to their home network and everything else works like it would without a vpn.
WireGuard routes certain traffic from the client (your iPhone) through the server (the computer at your house). If you route all traffic, then when your iPhone accesses the internet, it’s as if you were at home. Since that WireGuard server is sitting on your home LAN, it is able to route your phones traffic to anything else on that LAN, or out to the internet.
Wireguard clients have a setting called AllowedIPs that tells the client what IP subnets to route through the server. By default this is
0.0.0.0/0, ::/0, which means “all ipv4 and all ipv6 traffic”. But If all you want to access are services on your home LAN, then you change that to192.168.0.0/24or whatever your home subnet is, and only traffic heading to that network will be routed through the WireGuard server at your house, but all other traffic goes out of your phone’s normal network paths to the internet.Ahh. But what if you already used a VPN on the client for normal browsing etc - can you have two VPNs configured?
No, think of a VPN as a network cable. You can only send out of one or the other.
Now, if you are connected to a device that has another VPN to somewhere you want to go, then technically yes you would be using 2 VPN connections.
I don’t think iOS allows multiple VPNs to be enabled simultaneously. There appears to be only one VPN on/off toggle switch. From what I’ve seen you can have different vpn profiles but only enable one at a time. I could be wrong though.
Desktop operating systems like macOS, Linux (did I mention yet that I use arch Linux?), BSD, and um… that other one… oh yeah, Windows do allow this. I’m sure there are a variety of compatibility problems, but in general, multiple VPNs with the same or even different technologies can work together.
Okay, so that’s pretty much the setup I had in mind. Good to know there is not much need for an extra step for security, thanks for the answer !
Well, I guess that would still be vulnerable to DDOS attacks, but that would just prevent me from accessing my cinnamon apple-pie recipe from my self hosted recipe manager for some time. A bit mean, but not catastrophic.
I wondered if there would be some other attacks that could compromise my machine with only a wireguard setup, but that’s a good sign if there is nothing obvious.
You wouldn’t be any more vulnerable to ddos attacks than without WG.
Tailscale is easier than Wireguard but if you’re running OPNsense or OpenWRT it’s not hard to do a wireguard infra of your own and avoid having to use an outside service. I ended up having to revert to wireguard anyway because Tailscale’s android app wasn’t reliable on my new phone, it would drop out every few hours which messed up my monitor/alert system.
But Tailscale is still the easier of the two solutions.
Zero tier. I went tailscale originally, and they’re good, but their mdns support doesn’t exist and several services rely on it. (For me, the showstopper was time machine backups)
I like zerotier over wireguard because it’s one layer lower. So anything that uses Ethernet frames can be routed over it like it was a network switch plugged into your computer. This is probably why mdns works.
Do you test public WiFi with ZeroTier at all?
I ask because there’s a few public networks where WG won’t connect and I’m trying to find ways around it. I could always use cell data but this is more fun to me.
Yeah it’s worked everywhere I’ve tested. But that’s only really been airport WiFi, so I’m not sure it’s indicative of it working in general. It’s easy enough to setup for testing that it’s probably worth a shot
huh. I knew there was a reason for me to go back to ZT. mdns, you say? Nice to know!
Wireguard is the way to go. I like using wg-easy to use wireguard because its easy to set up in a docker container.
Agree, highly recommend this and/or Tailscale.
Double-pro. Running wireguard on docker assures that a native wireguard install won’t conflict with docker. Keep those tables in the same place.
I thought I would be the only one to try this. Would you share more details on your setup? I am interested because to me Wireguard is in the kernel so how could it be in a container.
Oh yeah. https://hub.docker.com/r/linuxserver/wireguard
Basically, docker can and does create network devices. It’s as easy for it to create wg0’s as it is to create networks for your other docker containers. If you’re going to run wireguard and docker, you’re better off to let docker handle the network routing and just run one of the various containers out there to stop them from fighting. That particular container is more general. You can run it client or server. Wg-easy, I believe is server-only, or even hide it inside other containers like docker-qbittorrent-wireguard, where it just hangs out and connects to whatever .conf you give it.
I did the whole thing in my early days selfhosting where I installed wireguard, docker, some apps, rebooted, everything breaks.
Install a wireguard container, configure it as you would, your apps, reboot… it still works, because docker isn’t conflicting with native wg-quick. It’s either this, or untangle and make an iptables setup permanent so when you reboot, it doesn’t break again.
Managing Wireguard is just one of Tailscale’s features.
It uses some UDP black magic fuckery to get through NAT and firewalls without having to open ports on the edge. Very useful if you’re behind CGNAT and/or your ISP is a dickhead and locks down the firewall on your router (this is why I use it; eat a dick, Vodafone). If the UDP fuckery is not available, it reverts to simple relay servers. The client can also advertise subnets and route to hosts on it. You could install the Tailscale client on OPNSense/pfSense/OpenWRT and access your entire home network through that one device.
