In previous posts we’ve covered everything required to set up a network with multiple VLANs and IPv6 (see part 1 for a list of all posts in this series). Today we’re going to talk about setting up an OpenVPN server on the ERL.
If you aren’t familiar with VPNs, think of them as encrypted tunnels used to connect computers over the internet. Common uses for VPNs include:
- Providing privacy while using public networks, especially open public wifi hotspots.
- Enabling secure, remote access to LANs, connecting remote networks. Corporations often use this to allow employees to access the corporate network remotely.
- Connecting networks in multiple, fixed locations. For example a business with multiple offices might use this to securely connect together the various office networks via the internet.
The VPN setup described here can be used for the first two use cases above but not for the third.
There are several protocols which can be used to set up a VPN, including PPTP, L2TP, SSTP, and OpenVPN. I’ve chosen OpenVPN here because it’s secure, flexible, and open source. The paticular configuration is very specific to my needs and level of paranioa. Consult the OpenVPN documentation to help you modify this setup for your needs.
OpenVPN uses public key cryptography in essentially the same way it’s used to make secure connections to websites. This means we need a public key infrastructure capable of generating signed public/private key pairs, which in turn means we need to create our own certificate authority (CA). There’s a script on the ERL to help us do that.
You’ll be asked a number of questions during the process, and skipping them may cause it to fail. Just answer the best you can. This is going to happen several times while setting up the certificates.
If you supply a password here, make sure you remember it. You’re going to need it! Of course if you forget it you can always start the process over again at the beginning.
Once this completes you’ll have a
demoCA directory containing a number of
files. The important ones are
cakey.pem, which is the private key for your
cacert.pem, which is the public key.
Next we’ll generate a public/private keypair for the OpenVPN server.
This will generate two files,
newcert.pem, which are
respectively the private and public halves of your server certificate.
Now we’re going to move all of these key files to
/config/auth/. This will
ensure that they’re preserved across firmware upgrades and include in your
Next we’ll generate a Diffie-Hellman parameter file. This will allow clients and the server to generate shared session keys without ever having to transmit that key over the internet, so even if someone compromised the server certificate they would be unable to decrypt session traffic. I use a key size of 2048, though 1024 is more common and probably safe (though I don’t claim to be a crypto expert).
This is going to take a while, so go get a cup of your favorite beverage. No need to hurry.
Once that’s completed you’ll need keys for clients, preferably one set per client.
At this point I note that some other guides state that you should remove the password from your client key files. I honestly can’t remember whether or not I did this, or maybe I just didn’t supply a password for the client certificates. Anyway, here’s the command to do this if you’re having issues because the client keys have a password, or if you’re just getting annoyed at entering the password each time you connect:
You’ll need to copy the
.pem files to the respective clients to allow them to
OpenVPN Server Setup
Now it’s time to set up the OpenVPN server on the ERL. This is done by creating a new interface. You’ll also need a new IPv4 subnet for the VPN; I use 192.168.200.0/24 here.
You’ll also need to make decisions about which port to use, whether to use tcp or udp, which routes to push, etc. For this example I’ll use tcp on port 443, which sort of disguises the traffic as normal SSL (useful because SSL connections are never blocked by network admins). I’ll push a route to allow communication with clients in the office network.
Update: As pointed out in the comments port 443 conflicts with using SSL for
the web gui. Your options to solve this are to either use a different port for
the web interface (the
serivce gui https-port setting controls this) or a
different port for the VPN.
Now we need to set up a firewall zone for the VPN and write rules for this zone. In this setup the VPN is really just an extension of the office LAN, so for the most part we can just reuse the same rules used for the office LAN zone. There are a handful of cases where this isn’t possible though:
- WAN to local: A rule is needed here to allow incoming tcp connections on port 443.
- VPN to office LAN: All traffic is allowed.
- Office LAN to VPN: All traffic is allowed.
Refer back to part 2 for help setting up the firewall.
Final Steps and Testing
The OpenVPN server hands out IP addresses to clients, so there’s no need to set up DHCP for the VPN subnet. You may or may not want to set up DNS to listen on the vtun0 interface, depending on your needs.
Now it’s time to save everything and try it out.
You’ll need to configure your client machine(s) to connect using the client certificate(s) we generated earlier. The steps for doing this vary depending on your OS; please consult the OpenVPN documentation for help.
For testing it’s ideal to try and connect from outside your network, e.g. by tethering to a phone. But for this setup it’s also possible to test by trying to use the VPN to access the office LAN from the home LAN. Just be sure you’ve set up firewall rules to allow clients on the home LAN to connect to the OpenVPN server on the router.
This section isn’t essential, but I do recommend it.
The OpenVPN hardening page covers various ways to improve the security of OpenVPN. It’s useful to read through these. The only one that I’m going to cover here is TLS auth.
The TLS auth option is pretty cool. It makes it so that the OpenVPN server will not respond to packets unless those packets have a valid signature from a pre-shared key. This makes it so that someone doing a port scan of your public IP address will not see that the OpenVPN port is accepting connections. It also provides other benefits described on the hardening page.
Setting this up is pretty simple. First you need to generate the pre-shared key.
Next, copy the key to
/config/auth on the ERL and to all client machines.
Then set up the OpenVPN server to use the
Finally, configure clients to pass the
--tls-auth ta.key 1 option to OpenVPN.
Since writing this post I’ve employed a few addtional hardening options for OpenVPN:
- Drop root privileges after OpenVPN initialization. This is done by passing the
--user nobody --group nogroupoptions to OpenVPN. Additionally the
--persist-key --persist-tunoptions should be used to avoid the need for privileges on soft restart.
- Use AES256 for the cipher and SHA256 for the message digest instead of the defaults (Blowfish/SHA1). Note that this may impact performance.
Use the following commands to enable these options.
You will also need to set the cipher and message digest appropriately in your client configuration.
In this post we’ve covered one fairly common scenario for setting up an OpenVPN server on the ERL. OpenVPN is incredibly flexible though, so if your needs aren’t completely covered by this guide there’s a pretty good chance that you just need to tweak the configuration.
The next post will wrap up this series by covering various useful odds and ends that we haven’t touched on yet.