The Hacker's Utility Belt: SSH
May 07, 2009
He's got strength and training on his side, but I could probably take him if he didn't have his belt and was blindfolded. Give him his utility belt though, and I'd be doomed. That's not because his belt has super powers—instead his belt contains tools that are useful in common situations of distress, and of course detective work. The Bat-grappling hook for instance allows him to scale buildings and walls, and tranquillizer darts allow him to temporarily disable a foe. Batarangs, with practice, can be used to disable an opponent by aiming for a body part, or to cut down a hanging chandelier, creating yet another obstacle for foes to fight through.
Hackers have similar tools, but we carry them in "/usr/bin/" (though how cool would it have been if the Bell Labs folks named it "/usr/belt/" instead). The tools we use, generally allow us to solve problems quickly and efficiently—just like Batman.
The tool that I've come to rely on quite a bit recently is SSH. Most hackers use SSH for working remotely on servers, and for copying files over SCP or SFTP. Some may even use FUSE and SSHFS to mount file systems over SSH. Its potential uses are endless, but the most useful uses for me lately have been tunneling and SOCKS proxies.
As a programmer who on occasion works from home (more so currently), it's often a battle to connect to all of the company resources I need in order to perform my job. Firewalls are good things, and are configurable to allow remote people access to walled gardens, but with dynamic IP addresses being the common norm, opening up resources to many different addresses gets hard for sysadmins to track. Besides, they've got a ton on their plate already.
To combat this problem, VPNs were created. Basically, a VPN allows one to connect networks together as if they were local to each other. When I use my companies VPN, it's as if I'm sitting in the office.
Often though, this isn't enough. When dealing with networks that clients own, they may provide us access through only one entry point, a single development server for instance, which is locked down.
This is where SSH comes in handy. In the OpenSSH implementation of the SSH client tools, the
-L option allows one to setup forwarding of traffic to another host. Suppose, I wanted to connect to a client's Oracle server, listening on port 1521:
$ ssh -N -L 9999:clients.oracle.server.name:1521 firstname.lastname@example.org
Now instead of pointing our SQL*Plus client to clients.oracle.server.name:1521, we point it to localhost:9999, and traffic is forwarded via your.companies.host as if we were connecting directly from your.companies.host.
This is extremely powerful, and simple.
Similar to the tunnelling example above, OpenSSH can act like a SOCKS proxy, allowing you to forward outbound traffic to a trusted source, which will then carry out the request on your behalf. This is great for browsing in public WI-FI spots where attackers might be sniffing for passwords being sent in plain text over HTTP, or some other unencrypted protocol.
And, of course, setting it up is as easy as passing the
-D option to ssh when connecting to you@remoteserver.
$ ssh -N -D 8000 user@remoteserver
Then, you can setup your computer (or web browser) to use a SOCKS proxy, and point it to localhost:8000.
This isn't without its problems though. For one, the network could block all outbound traffic except HTTP/HTTPS. One obvious workaround is to have a remote server listening for SSH connections on port 143 or port 80, instead of the protocol default of 22. But, you'll of course need your own box, or VPS for this, as your shared hosting account will probably not allow you to change sshd's listening port.
- I use the jargon file's definition of hacker, not the commonly confused term cracker.
- It can of course get much, much more complicated
- Unencrypted WIFI traffic is extremely easy to capture, with a tool like tcpdump
- The thought of a "loopback" SOCKS proxy, "loopback" tunnel or some combination of the two cannot work, as it would only ever forward traffic to your localhost's sshd, which would then forward traffic to the destination, unencrypted, over the local network—this is the exact situation we were trying to avoid. Plus, if you're trying to forward traffic in this way, over a port that's being blocked, it's still blocked.