Setting up a reverse SSH tunnel with a Raspberry Pi part 2

Notice! This post is more than a year old. It may be outdated.

Last time we got a Raspberry Pi set up and a VPS ready and accessible via SSH. If you have not read part 1 then I strongly encourage you to do that. In part 2 I promised to show an example use case on how to connect to a Windows Server over RDP via a reverse SSH tunnel.

Testing SSH access

First, let’s make sure that the reverse SSH tunnel is working as expected. I’m going to create an SSH session to the VPS and from there I’m going to create another SSH session to the Pi.

Connecting to the external server:

$ ssh user@serverhostname

From there let’s create another SSH session to our Pi which is inside the office network:

$ ssh pi@localhost -p 2222

Why connect to localhost on port 2222? Recall from part 1 that we created a reverse SSH tunnel from the Pi to the server. We defined that all connections from our server to localhost on port 2222 should be forwarded to the Pi.

If you were able to connect to your Pi then the tunnel is working.

Tunneling RDP traffic

For the sake of the example I’m going to assume the IP of the Windows Server at the office network is 192.168.1.10 and the server outside of the network can be accessed via myserver.com. Two SSH tunnels are needed to forward traffic from the client machine to myserver.com and from there to the Pi.

External server -> Raspberry Pi -> Windows Server

First of all, I’m going to create a tunnel to forward RDP traffic from myserver.com to the Windows Server.

$ ssh -L 3000:192.168.1.10:3389 pi@localhost -p 2222

This forwards all connections made to localhost on port 3000 to the Windows Server (192.168.1.10) using port 3389. -L flag specifies that the given port on the local (client) host is to be forwarded to the given host and port on the remote side. It is achieved by creating an SSH tunnel to the Pi via a reverse SSH tunnel we set up previously. Note that pi@localhost -p 2222 in the command is the same used previously to test the SSH tunnel.

Client machine -> External server

From the client machine an SSH tunnel must be created to forward all connections made to localhost port 5000 (just a random port I chose) to the external server.

$ ssh -L 5000:localhost:3000 user@myserver.com

Now when you connect your RDP client to localhost:5000, this connection is forwarded to myserver.com and from there a new connection is made to localhost:3000. Looking at the preceding step, you can see that this will be in turn forwarded the Windows Server.

Recap

In short, the following steps were done:

From the Pi create a reverse SSH tunnel to an outside server.
From an outside server tunnel incoming traffic to the Pi and from there to the Windows Server.
From the client machine tunnel RDP traffic to an outside server.

Summary

There are a couple of more things you need to do to make it a reliable solution. Obviously the steps need to be automated. It can get cumbersome to create two SSH tunnels over and over again. Additionally the reverse SSH tunnel needs to be kept alive. Network errors and power outages can occur which will kill the tunnel. This means you have to physically go to the Pi and redo the tunnel. I would recommend creating a cronjob which checks if the tunnel is up and recreates it if it is not.