Skip to content
← All writing

· 9 min read

Site-to-Site VPNs with WireGuard and pfSense: A Practical Guide

  • Security
  • Networking
  • WireGuard
  • pfSense
  • AWS

When you need a site-to-site VPN#

A site-to-site VPN joins two networks so that machines on either side can talk as if they were on one private network. Typical cases:

  • On-premise devices (for example IP cameras streaming RTSP) that a cloud service must reach without exposing them to the internet
  • A lab or office network that needs private access to services inside a cloud VPC
  • Two cloud environments in different providers or regions

I have configured site-to-site VPNs in production and built RTSP camera-streaming pipelines, which is where these two topics usually meet. This guide uses WireGuard on pfSense at the remote site and a Linux WireGuard gateway inside an AWS VPC, but the principles carry over to any pair of endpoints.

Why WireGuard#

  • Small and auditable: a few thousand lines of code instead of the large codebases of older VPN stacks.
  • Modern cryptography by default: no cipher negotiation to misconfigure.
  • Fast reconnects: roaming and NAT changes are handled gracefully.
  • Simple model: every peer is a public key plus the IP ranges it is allowed to use.

Step 1: plan the addressing#

Overlapping subnets are the number-one cause of broken site-to-site VPNs. Decide the ranges first:

Network CIDR
Remote site LAN (behind pfSense) 192.168.50.0/24
AWS VPC 10.0.0.0/16
Tunnel network 10.99.0.0/30

The tunnel network is a tiny point-to-point range used only for the two WireGuard interfaces.

Step 2: the cloud gateway#

Inside the VPC, a small Linux instance in a public subnet acts as the gateway. Two AWS settings are easy to forget:

  1. Disable source/destination check on the instance, otherwise AWS drops traffic it forwards.
  2. Add a route in the private subnets' route table: 192.168.50.0/24 → the gateway instance.

WireGuard config on the gateway (/etc/wireguard/wg0.conf):

[Interface]
Address = 10.99.0.1/30
ListenPort = 51820
PrivateKey = <gateway-private-key>
PostUp = sysctl -w net.ipv4.ip_forward=1

[Peer]
# pfSense at the remote site
PublicKey = <pfsense-public-key>
AllowedIPs = 10.99.0.2/32, 192.168.50.0/24

The security group only needs UDP 51820, and ideally only from the remote site's public IP.

Step 3: pfSense at the remote site#

On pfSense (with the WireGuard package installed):

  1. VPN → WireGuard → Tunnels: create a tunnel, generate keys, listen port 51820.
  2. Peers: add the cloud gateway with its public key, endpoint <elastic-ip>:51820, keepalive 25 seconds, and allowed IPs 10.99.0.1/32, 10.0.0.0/16.
  3. Interfaces → Assignments: assign the WireGuard tunnel as an interface, give it 10.99.0.2/30, and enable it.
  4. System → Routing: add a gateway at 10.99.0.1 on that interface and a static route for 10.0.0.0/16 through it.

The persistent keepalive matters when pfSense sits behind NAT: it keeps the mapping open so the cloud side can always reach it.

Step 4: firewall rules (least privilege)#

A tunnel that is up is not the same as a tunnel that is safe. Allow only what the use case needs.

On the pfSense WireGuard interface, for a camera use case:

Action Source Destination Port
Pass 10.0.0.0/16 (streaming service subnet only) camera VLAN TCP/UDP 554 (RTSP)
Block any any any

On AWS, the security groups of the services should accept traffic from 192.168.50.0/24 only on the ports they serve. Never open the whole VPC to the remote site "temporarily".

Step 5: verify and monitor#

# On the gateway: is the handshake recent? Are bytes flowing?
sudo wg show
# Path check from the cloud side
ping -c3 192.168.50.10
# Can the service reach the camera stream?
ffprobe rtsp://192.168.50.10:554/stream1

A handshake older than a few minutes means the tunnel is down. Exporting wg show data to your monitoring and alerting on handshake age catches outages before users do.

Hardening checklist#

  • Restrict the listening port to known peer IPs wherever possible.
  • Rotate keys when people with access leave, and keep private keys out of Git.
  • Segment the remote site so the tunnel only reaches the VLANs it needs.
  • Log blocked traffic on both ends during rollout; it reveals what is really being used.
  • Document addressing and rules next to the code that creates them (Terraform for AWS, config backups for pfSense).

Summary#

A WireGuard site-to-site VPN is one of the simplest ways to connect on-premise devices to cloud services without exposing them publicly. Plan addresses first, remember the AWS routing details, and spend most of your effort on firewall rules: the tunnel is the easy part, least privilege is the important part.