Re: [secdir] Secdir review of draft-ietf-dcrup-dkim-usage-04

["Salz, Rich" <[e-mail]>]

Because of limitations in the DNS system, RSA 1K is really all that’s practical.  For stronger security we are moving to ECC, which is covered in a separate I-D.  This document is really about getting rid of RSA-512.


From: Magnus Nyström <[e-mail]>
Date: Tuesday, September 12, 2017 at 11:43 PM
To: Russ Housley <[e-mail]>
Cc: "[e-mail]" <[e-mail]>, "[e-mail]" <[e-mail]>
Subject: Re: [secdir] Secdir review of draft-ietf-dcrup-dkim-usage-04

Yes, Russ. Perhaps "When using RSA, signers MUST use keys at least 2048 bits long." (I could imagine that at some point DKIM would support ECC)
Thanks,

On Tue, Sep 12, 2017 at 8:11 PM, Russ Housley <[e-mail]<mailto:[e-mail]>> wrote:
Magnus:

I agree with your recommendation, but I would like to make sure I understand your suggestion.  The document says:

   ...  Since short RSA keys more easily succumb to
   off-line attacks, Signers MUST use RSA keys of at least 1024 bits for
   all keys.  Signers SHOULD use RSA keys of at least 2048 bits. ...

You want to change "1024" to "2048", and then drop the following sentence, right?

Russ


On Sep 12, 2017, at 2:33 AM, Magnus Nyström <[e-mail]<mailto:[e-mail]>> wrote:

I have reviewed this document as part of the security directorate's ongoing effort to review all IETF documents being processed by the IESG. These comments were written primarily for the benefit of the security area directors. Document editors and WG chairs should treat these comments just like any other last call comments.

This document intends to update the DKIM specification with a new mandatory hash algorithm (SHA-256) and new RSA key size requirements.
While I definitely agree with the stated direction, I do wonder about the RSA 1024 bit key size recommendation. Conventionally, this corresponds to about 80-bit security and to reach the equivalent of 128-bit security (which is what SHA-256 gives), a 3072-bit RSA key size should be recommended. In this day and age, mandating only 1024 bits seems a little weak. I recognize there may be limitations in the DNS records storing these keys, but it should be possible to store at  least 2048-bit keys (256 bits) (corresponding roughly to 112-bit security) or at least close to it and thus why not require 2048 bit RSA keys as a minimum? 1024 bit keys are, as is also commonly known, considered "legacy" by NIST SP 800-57 part 1 and shouldn't be used for new signatures at this point.

-- Magnus




--
-- Magnus