ssh no ports provides a way to ssh to a remote linux host/device without that device having any open ports (not even 22) on external interfaces. All network connectivity is out bound and there is no need to know the IP address the device has been given. As long as the device has an IP address, DNS and Internet access, you will be able to connect to it.
There are two binaries:-
sshnpd : The daemon that runs on the remote device
sshnp : The client that sets up a connection to the device which you
can then ssh to via your localhost interface
To get going you just need two atSigns and their .atKeys files and the
binaries (from latest
It's also possible to run from the source here using
dart run. Once you have
the atSigns (free or paid atSigns from atsign.com), drop the binaries in place
on each machine and put the keys in
~/.atsign/keys directory. You will need
a device atSign and a manager atSign, but each device can also have a unique
device name using the --device argument.
Once in place you can start up the daemon first on the remote device. Remember to start the daemon on start up using rc.local script or similar.
./sshnpd --atsign <@your_devices_atsign> --manager <@your_manager_atsign> \ --device <iot_device_name> -u
Once that has started up you can run the client code from another machine.
./sshnp --from <@your_manager_atsign> --to <@your_devices_atsign> \ --host <example.com> -l --local-port --device <iot_device_name>
The --host specifies a DNS name of the openssh server of the client machine that the remote device can connect to. If everything goes to plan the client will complete and tell you how to connect to the remote host for example.
ssh -p 3456 [email protected]
When you run this you will be connect to the remote machine via a reverse ssh tunnel from the remote device. Which means you can now turn off ssh from listening all all interfaces instead have ssh listen just on 127.0.0.1.
That is easily done by editing
#Port 22 #AddressFamily any ListenAddress 127.0.0.1 #ListenAddress ::
And restarting the ssh daemon. Please make sure you start the sshnpd on startup and reboot and check.. As this is beta code it is suggested to wrap the daemon in a shell script or have sysctld make sure it is running.
My preference whilst testing was to run the daemon in TMUX so it is easy to see the logs (-v).
Thoughts/bugs/contributions via PR all very welcome!
Run the daemon binary file or the dart file:
dart run bin/sshnpd.dart <args|flags>
|--keyFile||-k||false||Sending atSign's keyFile if not in
|--atsign||-a||true||atSign of this device|
|--manager||-m||true||Manager's atSign, that this device will accept triggers from|
|--device||-d||false||Send a trigger to this device, allows multiple devices share an atSign||"default"|
|--[no-]sshpublickey||-s||Update authorized_keys to include public key from sshnp|
|--[no-]username||-u||Send username to the manager to allow sshnp to display username in command line|
The daemon can also be deployed as part of a pre built docker container,
that also has a number of networking tools installed. The container image
is located on Dockerhub as
atsigncompany/sshnpd:latest or you can build
your own using the Dockerfile in the root of the project.
The image expects to have the atKeys for the atSign being used in the
/atsign/.atsign/keys directory, this can be mounted as a volume at startup
of the docker run command using
-v $(pwd):/atsign/.atsign/keys/ assuming
you are in the dircetory where the atKeys file is located. The full comand
to start the container would be something like this:-
docker run -v <location of atKeys>:/atsign/.atsign/keys/ atsigncompany/sshnpd "-a <atSign> -m <atSign> -d <device name> -v -u"
Once the container is running to log into the container the sshnp command would be used as normal, but you will log into the container not the host, from the container you could then log into the host or any other local network hosts you have access to.
Docker is very well documented and if you want to keep the container running after a reboot if for some reason the container crashes is all easily achieved.
Run the binary file or the dart file:
dart run bin/sshnp.dart <args|flags>
|--key-file||-k||false||Sending atSign's atKeys file if not in
|--to||-t||true||Send a notification to this atSign|
|--device||-d||false||Send a notification to this device||"default"|
|--host||-h||true||FQDN Hostname e.g. example.com or IP address to connect back to|
|--port||-p||false||TCP port to connect back to||22|
|--local-port||-l||false||Reverse ssh port to listen on, on your local machine||2222|
|--ssh-public-key||-s||false||Public key file from
|--local-ssh-options||-o||false||Add these commands to the local ssh command, useful if you want to tunnel ports|
The instructions above work for a system where the person doing the admin of the machine connected to by sshnp is able to run an SSH daemon that's open to the Internet. But that's often not practical for many of the same reasons why the device can't/won't be reachable directly with an open port. To get around this issue it's possible to use the Ngrok service as a proxy for the inbound SSH connection.
From their signup page
From the system you're using for admin:
Then copy the key and paste it into the
New SSH Key box on the
SSH Public Keys page.
Such as OpenSSH. It can run on any port, and only needs to be bound to
localhost. The following example illustrates the use of an SSH server
bound to port 2222. So the example
/etc/ssh/sshd_config above becomes:
Port 2222 #AddressFamily any ListenAddress 127.0.0.1 #ListenAddress ::
It may be useful to do this in a
as another terminal will be needed for
ssh -R 0:localhost:2222 tunnel.us.ngrok.com tcp
This will initialise a connection showing something like:
Allocated port 12357 for remote forward to localhost:2222 ngrok (via SSH) (Ctrl+C to quit) Account Demo McDemoname (Plan: Free) Region us Forwarding tcp://6.tcp.ngrok.io:12345
Command line form:
sshnp -f <@your_manager_atsign> -t <@your_devices_atsign> \ --device <iot_device_name> -h 6.tcp.ngrok.io -p 12345 -l 3456
NB: Ngrok is likely to provide a different tunnel server and port
each time. So substitute the values from the actual connection for
-h 6.tcp.ngrok.io and
sshnp -f @happyadmin -t @moresecurething \ --device demothing -h 4.tcp.ngrok.io -p 10646 -l 3456
Connect to it with something like:
ssh -p 3456 -i ~/.ssh/key_for_device.key [email protected]
-p 3456corresponds to
-l 3456from the
-i ~/.ssh/key_for_device.keyis presenting a private key that's trusted by the device in its
deviceuseris the username for the device
First a tunnel from Ngrok back to admin_PC:
ssh -R 0:localhost:2222 tunnel.us.ngrok.com tcp admin_PC Ngrok 2222<-------------------12345 <----------------------------
Then a tunnel initiated by
sshnp from the device, through Ngrok to the
sshnp -f @happyadmin -t @moresecurething \ --device demothing -h 0.tcp.ngrok.io -p 12345 -l 3456 admin_PC Ngrok admin_PC 2222<-------------------12345 Device 3456<----/‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾\-------22 <------------------------------------------------ <----------------------------
Finally an SSH connection through those tunnels from the admin_PC to the device:
ssh -p 3456 -i ~/.ssh/key_for_device.key [email protected] admin_PC Ngrok admin_PC 2222<-------------------12345 Device SSH------>3456<----/‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾\-------22-------->SSHD \______________________________________________/ <------------------------------------------------ <----------------------------
Of course that final SSH connection can also be used as a tunnel...
Anyone running ssh where they don't want it to be open to a hostile network!
Created by Atsign
Original code by @cconstab