Update README.md
Add license
Update README.md
Running this script will do two things:
This script takes no arguments and will ask you for input interactively.
Why is this useful? Why don't we just use SSH?
The main reason is... firewalls: most firewalls only allow HTTP/HTTPS outbound traffic. Typically people worked around this by configuring their SSH servers to listen on port 443 (the HTTPS port). However, most firewalls nowadays do not just look at the destination port, but can actually tell if the outgoing traffic is TLS (which is where HTTP is encapsulated in HTTPS) or not.
Using a TLS tunnel instead of an SSH one makes it harder (but not impossible!) for firewalls to detect if this is HTTPS traffic or not.
NOTE:
It is not imposible because HTTP traffic is much less 'interactive' than
the one generated from a remote shell, and an intelligent enough firewall
should be able to detect this (however, so far, I don't think I have come
across any of these)
Are there any drawbacks? Sure: speed. The channel is being encrypted twice (once by TLS and another one by SSH). Why do we use SSH, then? Couldn't we just directly open a shell inside the TLS tunnel instead of running SSH? Yes, we could... but then we would miss all the benefits SSH brings with it: users management, redirection tricks, etc...
NOTE:
Also, the "speed reduction" is minimal. I haven't been able to notice it.
You must run this script from a computer that has SSH access to the server you want to configure on port 22. This is only needed at this stage. Later, for the client to access the server, only port 443 needs to be reachable from the Internet.
The server must not be using port 443 (ie. you cannot have a web server already running there). If this is not acceptable, forget about this script and try following the instructions detailed on this blog post:
http://blog.chmd.fr/ssh-over-ssl-episode-4-a-haproxy-based-configuration.html
This script is just a wrapper that calls 'openssl' (to generate the certificates) and 'socat' (to create the tunnel). You need both of them installed on the client and the server (and, obviously, 'ssh'/'sshd'):
Examples:
Debian/Ubuntu --> $ sudo aptitude install socat openssl
Arch --> $ sudo pacman -S socat openssl
First, using openssl, a pair of TLS (X509) certificates will be generated (one for the server and one for the client)
NOTE:
Every time this script runs, different pairs of certificates will be
generated. This means you cannot run this script twice and expect the
server configured with the first instance to accept connections from the
client script created in the second instance.
Then, a remote server will be configured (using SSH) like this:
The server certificate and the public part of the client certificate are copied to the '/etc/tls_tunnel_server' folder.
The init system (either 'init.d' or 'systemd') is configured to run a 'socat' command that, using the just copied certificates, creates a TLS tunnel that is connected to localhost:22 (ie. the SSH server). Only a client with the corresponding client certificate will be granted access to the SSH server behind.
Finally, a script is generated (meant to be used by the client). This script (a single 'bash' script) contains everything the client needs: the client certificate, the public part of the server certificate and the logic to run socat to establish the TLS tunnel and SSH to use it.
The generated certificates are valid for one year, but you can easily change
this: search for -days 365
in the script and set the new desired
value.
The generated client script ("ssh_over_tls_tunnel_client.sh") will only connect to the server if it presents the expected certificate. This means that, as in a regular HTTPS connections, you can be sure there is no man in the middle.
The server will only accept connections from clients with the expected certificate. This certificate is "embeded" inside the generated client script ("ssh_over_tls_tunnel_client.sh") in plain text. Because of this you should probably not upload this script to "the cloud". Fortunately, in addition, an encrypted version of this same file is also generated: "ssh_over_tls_tunnel_client.sh.enc.b64.sh". When executed it will ask you for the original encryption password and then run. It is safe to upload this to public servers.
NOTE:
Even if you don't encrypt it, the amount of damage that results from
this script being leaked to a third party is minimal as it only grants
the owner permission to establish connection to the SSH server, which
requires authentication anyway.