Re: [TLS] Encrypt-then-MAC again (was Re: padding bug)

[[e-mail] (Martin Rex)]

Peter Gutmann wrote:
>
>Michael D'Errico <[e-mail]> writes:
>> 
>>In trying to figure out what's stalling the encrypt-then-mac draft
> 
> What's stalling the draft is that the WG chairs refuse to accept it, and
> nothing else.  This is a simple, trivial fix for a decade-old problem that's
> come back to bite us again and again, that several implementations have
> already implemented, and that a straw poll on the list a while back indicated
> had general support.  In other words the requirements for rough consensus and
> running code was met some time ago.

I've been thinking over this proposal for quite a while, and I've made
up my mind.  I'm now strongly opposed.

I'm perfectly OK with a solution for fixing the mac-pad-encrypt goof
in GenericBlockCipher PDU for all existing versions of TLS, but I'm
strongly opposed to moving the HMAC into the clear, and into particular
I am strongly opposed to put the HMAC into the clear for the
GenericStreamCipher PDU.

I'm going to prepare a proposal I-D to do exactly (and only) what
Serge Vaudenay had recommended in 2001, and what TLSv1.1 failed to fix,
change mac-pad-encrypt into pad-mac-encrypt for the GenericBlockCipher PDU.
When doing this, we can also consistently adopt the explicit random IV
(which can be implemented as a random confounder) to all versions of TLS.

Signaling of this PDU change will be through a new SCSV in ClientHello,
confirmation by the Server through a simple ServerHelloExtension, similar
to how signaling is done for rfc5746 (TLS renegotiation_info extension).

-Martin