[jdev] Re: TLS and self-signed certs

[Peter Saint-Andre]

In article <200411111955.28246.neil at hakubi.us>,
 Neil Stevens <neil at hakubi.us> wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> On Thursday 11 November 2004 05:06 pm, Justin Karneges wrote:
> > While JD's comments sum this up nicely, I just want to reiterate loudly
> > that self-signed certificates alone truly are worthless.  I'm not even
> > talking about man in the middle attacks either.  As a form of identity,
> > a self-signed cert is as effective as the "From:" header in good old
> > SMTP, and this would allow spammers to get right in and start faking
> > domains.
> 
> Wrong.  If a certificate remains unchanged, then you know that as long as 
> it is unchanged, you're continuing to connect to the server you connected 
> to in the past.
> 
> You can't know if there's a man-in-the-middle in progress when you first 
> connect, but if you're remembering certificate and someone tries one after 
> a while, you will be able to detect that.
> 
> ssh does this, for example.

Precisely. And one can argue that ssh is the most-used encryption 
technology on the planet. Perhaps "opportunistic cryptography" is not a 
bad model to follow? Even the IETF seems to be moving in the direction 
of recognizing reality on this issue -- see the "Better Than Nothing 
Security" BOF at IETF 61 this week:

http://www.financialcryptography.com/mt/archives/000247.html#more

Peter