[lamps] Re: Call for adoption of draft-davidben-x509-alg-none

[Tim Hollebeek <[e-mail]>]

Yes, after thinking about it a bit, I think the adjective I like the most is “unsigned”. This compares nicely with “self-signed”, which “seems like it is signed but isn’t”. Upgrading to “HEY, THIS ISN’T SIGNED BY ANYONE” is an improvement over that.

 

This is the “this page intentionally left blank” of signatures, that makes it very clear the cert is not signed by anything. That is not the same as saying it can’t be trusted.

 

Just that signature verification should fail, because … it is explicitly unsigned.

 

-Tim

 

From: David Benjamin <[e-mail]> 
Sent: Thursday, February 27, 2025 1:11 PM
To: Salz, Rich <rsalz=[e-mail]>
Cc: Andrew Chen <[e-mail]>; Deirdre Connolly <[e-mail]>; David von Oheimb <it=[e-mail]>; LAMPS <[e-mail]>
Subject: [lamps] Re: Call for adoption of draft-davidben-x509-alg-none

 

I think there's a misunderstanding here about what this is. Did you see the presentation at the interim? There might be a record.

 

Signature-less certificates are no more insecure or unsafe than a self-signed certificate. In both cases, the signature (or lack thereof) gives you no assurance. In order to care about a signature, you have to care about the public key. The self-signature comes from the same public key that it is signing over! It should not help you believe in the information in any way.


Rather, signature-less or self-signed, these are used in contexts where you have out-of-band information about the key, etc., in the certificate. This is not some niche use case. Every X.509 validation starts with out-of-band information: the list of trust anchors that the relying party accepts.

 

What we're dealing with is a chicken-and-egg problem. A certificate is an edge between two nodes in the PKI, issuer -> subject. Trust anchors and other cases of out-of-band information are nodes. They just want information about some subject. Unfortunately, in the long history of X.509, the world did not evolve into giving those a different structure, but to just use self-signatures as a syntactic hack to make edges look like nodes, even though the signature doesn't do anything useful. Since we've gotten ourselves into that space, this provides a less inefficient (and problematic w.r.t. cross-protocol issues) version of that syntactic hack.

 

On Thu, Feb 27, 2025 at 1:02 PM Salz, Rich <rsalz=[e-mail] <mailto:[e-mail]> > wrote:

ZjQcmQRYFpfptBannerEnd

I'm thinking of the eventual future where these supplant self-signed certs. Someone who stumbles upon one of these (e.g. in sample/example code included with a software package) should know that they're insecure/unsafe and not meant for production use.

 

But that’s not they point. They *could* be used for production use and the trust decision must be out-of-band, which is probably the thing to call them.

_______________________________________________
Spasm mailing list -- [e-mail] <mailto:[e-mail]> 
To unsubscribe send an email to [e-mail] <mailto:[e-mail]>