summaryrefslogtreecommitdiff
path: root/website/why.mdwn
blob: 7f69614b27a7cd17889fac4ebd15fcf60056860b (plain)

[[meta title="Why should you be interested in the MonkeySphere?"]]

Why should you be interested in the MonkeySphere?

As an ssh user

Do you use ssh to connect to remote machines? Are you tired of seeing messages like this?

The authenticity of host 'foo.example.org (192.0.2.3)' can't be established.
RSA key fingerprint is 17:f4:2b:22:90:d4:98:9a:a2:c5:95:4e:4a:89:be:90.
Are you sure you want to continue connecting (yes/no)?

Do you actually tediously check the fingerprint against a cryptographically-signed message from the admin, or do you just cross your fingers and type "yes"? Do you wish there was a better way to do it? Shouldn't our tools be able to figure this out automatically?

Do you use ssh's public key authentication for convenience and/or added security? Have you ever worried about what might happen if you lose control of your key? (Or did you have a key that was compromised by the OpenSSL debacle?) How many accounts/machines would you need to clean up to ensure that your old, bad key is no longer in use?

Have you ever wished you could phase out an old key and start using a new one without having to comb through every single account you have ever connected to?

As an sshd administrator

If you are a system administrator, have you ever tried to re-key an SSH server? How did you ease the change along to your users? How did you keep them from getting the big scary warning messages?

Have you ever wanted to allow a colleague key-based access to a machine, without needing to have a copy of their public key on hand?

Have you ever wanted to be able to revoke the ability of a key to authenticate across the entire infrastructure you manage, without touching each host by hand?

What's the connection?

These questions all stem from rough edges we run up against in regular use of SSH that could be improved by a decent Public Key Infrastructure (or PKI). A PKI at its core is a mechanism to provide answers to a few basic questions:

  • Do we know who a key actually belongs to? How do we know?
  • Is the key still valid for use?

Given a clearly stated set of initial assumptions, functional cryptographic tools, and a PKI, these questions can be clearly answered in an automated fashion. We should not need to ask humans to do complicated, error-prone things (e.g. checking host key fingerprints) except in relatively rare situations (e.g. when two people meet in person for the first time).

The good news is that this is all possible, and available with free tools!

Examples

Bob is an ssh user, and has just been given an account on foo.example.org by Alice, the example.org system administrator, who he knows.

Bob already trusts Alice to properly identify all example.org servers. Alice already knows who Bob is, and the new machine foo knows that it can rely on Alice's certifications because Alice is its administrator.

Alice can set up the new bob account on foo.example.org without needing to give Bob a new passphrase to remember, and without needing to even know Bob's current SSH key. She simply tells foo that Bob <bob@example.net> should have access to the bob account.

Bob's first connection to his new bob account on foo.example.org is seamless, because all the steps are already in place! Using the MonkeySphere, Bob never has to "accept" an unintelligible host key or type a password.

When Bob decides to change the key he uses for SSH authentication, he can do so at once: he generates a new key, revokes his old key, and publishes these changes to the public keyservers. The next time he's ready to log into foo.example.org, it accepts his new key -- and it won't accept his old key any longer.

The same thing works for Alice when she decides to re-key foo.example.org (let's say Alice learned that Eve has compromised the old key). Alice generates a new key, revokes the old one, publishes the changes, and the next time Bob connects, he connects as smoothly as ever. And if Eve tries to use the old host key to masquerade as foo, Bob's SSH client will refuse to let him connect!

Alice can even quit as example.org system administrator, and revoke her certifications of all example.org hosts. As long as Bob knows and trusts the new example.org system administrator to identify hosts in that domain, there's no problem.

Why OpenPGP?

We believe that OpenPGP is the right PKI to use for this project. It allows a very flexible trust model, ranging all over the map, at the choice of the user:

  • individual per-host certifications by each client (much like the stock OpenSSH behavior),

  • strict centralized Certificate Authorities (much like proposed X.509 models), and

  • a more human-centric model that recognizes individual differences in ranges of trust and acceptance.

Even if Bob doesn't trust Alice to identify all example.org hosts, his first connection to foo.example.org should give him more than an unintelligible string to accept or reject. It should also give him the information that Alice (and perhaps her colleague Charles) have certified the key. This is far more useful information than the current infrastructure allows, and is more meaningful to actual humans using these tools than some message like "Certified by GloboTrust".