In a reply to Paul on [1], and Tony on [metzdowd, awaiting mod approval], I 
detailed a realization that Certificate Transparency (CT), does not seem to 
detect certificate mis-issuance even if clients are able to successfully gossip 
Signed Tree Heads (STHs) as per [2].

Steve Kent noticed this problem [3] (see explanation below) in his OP to [1], 
although he didn't investigate whether gossip would fix it or not.

[1] http://www.ietf.org/mail-archive/web/trans/current/msg00588.html
[2] http://www.ietf.org/proceedings/90/slides/slides-90-trans-2.pdf
[3] http://www.ietf.org/mail-archive/web/trans/current/msg00534.html

Why gossip doesn't work.
===================

The realization is that a rouge CA can do its thing and generate a fraudulent 
cert to MITM connections *and* log this cert to a log *and* still get away with 
MITM undetected even with clients gossiping STHs.

STHs are used to generate consistency proofs, but consistency proofs just tell 
you that you're the log you're currently looking at includes everything what 
you saw previously (from an older Merkle tree).

The audit proofs just tell you that a certificate exists in a log. It doesn't 
tell you that a certificate is fraudulent. There's nothing preventing 
fraudulent certs from existing in a log. The proof will tell you it's there, 
but it won't tell you it's fraudulent.

For more on how these proofs work, see Google's documentation here:

http://www.certificate-transparency.org/log-proofs-work

So now, Clog MITM's millions of users, they gossip STHs, it stops MITMing them, 
they see the original cert, they exchange the same STHs (which don't change 
just because clients see the original cert).

The only thing that could detect this are the Monitors, but they aren't going 
to save you because they would need to monitor *all* logs for *all* domains and 
alert *everyone* about those changes (difficult enough, you'd need a... 
b-b-b-blockchain) and get those alerts safely (without censorship or tampering) 
to everyone on Earth.


Game Over for Certificate Transparency?
================================

Unless I'm mistaken, this seems to be Game Over.

Given that another human being also independently saw this problem [3], I am 
feeling confident enough to share this on this list.

Maybe the blockchain can save CT, but if it does, people will realize that they 
don't need Certificate Transparency, so it seems like game over either way.

Cheers,
Greg Slepak

--
Please do not email me anything that you are not comfortable also sharing with 
the NSA.

Attachment: signature.asc
Description: Message signed with OpenPGP using GPGMail

_______________________________________________
Messaging mailing list
[email protected]
https://moderncrypto.org/mailman/listinfo/messaging

Reply via email to