In software engineering, it is important to remember the principle of redundancy. Not because it is useful for software, but because it is useful for humans.
Human beings work continuously with redundancy because most human information processing is soft and fuzzy. How does a system deal with soft, fuzzy results? It takes readings from different sources, as little correlated as possible, and compares them. If three readings from independent sources all suggest the same conclusion, then we are good. If 2 out of 3 say good, then the human brain says "take care," and if 1 out 3 is good, then it is discarded.
In comments on the last post, Peter G explored the direct question of whether anyone checked the fingerprint of the SSH server:
I tried to get some data a while back on SSH key checking in response to SSH fuzzy fingerprints (if you're not familiar with fuzzy fingerprints, they create a close-enough fingerprint to the actual target to pass muster in most cases). Because human-subject experimentation requires a lot of paperwork and scrutiny, I thought I'd first try and establish a base rate for SSH fingerprint checking in general. In other words if you set up a new server with a totally different key from the current one, how many people will be deterred?
So I tried to establish the fingerprint-check rate in a population of maybe a few thousand users.
It was zero.
No-one had ever performed an out-of-band check of an SSH fingerprint when it changed.
Given a base rate of zero, I didn't consider it worthwhile doing the fuzzy fingerprint check :-).
What is going on here? Three things. For some reason that has never been explained, SSH has never made it easy to check the fingerprint. Like OpenPGP to some extent, the fingerprints have been delivered in incompatible formats across different channels. E.g., my known_hosts file says that a server I know is AAAAB3N.... and the only way to easily see the fingerprint is to simulate a compromise by clearing the cache.
Secondly, and the point of this post: the fingerprint is only one of the datums that is being displayed. In the last post, I talked about two other pieces of data: One was the failure of server-key matching, and the other was the fallback to password-request.
The key lesson here is that SSH delivers enough information to do the job: it isn't the fingerprint per se, but the whole package of fingerprint, server-key matching, and precise mode.
Three data points, albeit rather poorly presented. Which brings us to another point: In practice, this is only good enough in rare and experienced situations. That breach was only picked up because of the circumstances and a good dose of luck.
This leads us to conclude that SSH is only just good enough, sometimes. Why? Because it is only just good enough for the job; it's circular, because since it has done the job well enough all these years, the security model has not been much improved against the theoretical concept that knocked the theoretical MITM on the head. The third thing is then lack of attacks -- now, however, circumstances are changing, and improvements should take place. (Indeed, if you have a longer perspective, you'll notice that the distros of SSH have been upgrading the security model over the last few years.)
But, the important upgrades do not want to be in forcing the fingerprint down the throats of the user. Instead, they want to be in the area of redundancy: more uncorrelated soft and fuzzy signals to the user, that work with the brain, not with the old 1990s computer textbooks.
Hence, and to complete the response to Peter G, this is why the PKI apologists are looking in the wrong area:
So there is a form of data available, but because it's not very interesting it'll never be written up in a conference paper (there's a longer discussion of fuzzy fingerprints and related stuff in my perpetual-work-in-progress http://www.cs.auckland.ac.nz/~pgut001/pubs/usability.pdf). I've seen this type of authentication referred to as leap-of-faith authentication in a few recent usability papers, and that seems to be a reasonable name for it. That's not saying it's a bad thing, just that you have to know what it is you're getting for your money.
Yeah, we can all see "leap of faith" as a sort of mental trick to avoid really examining why SSH works and why PKI doesn't or didn't. That is, "oh, but because you make this 'leap of faith' you're not really secure, according to our models. So I don't have to think any more about it."
The real issue here is again, it worked, enough, for the job. Now the SSH people will think more, and upgrade it, because it is being attacked. I hope, at least!
The PKI people cannot say that. What they can say is "use TLS/SRP" or some other similar RFC acrophiliac verbage which doesn't translate to anything a user can eat or drink. Hence, the simple answer is, "come back when I can use it."Posted by iang at February 20, 2008 02:48 PM | TrackBack