Re: [Dcrup] rsa-sha1 usage

When looking at the global usage of SHA-1, it's important to note that
several of the largest email service providers still use rsa-sha1
signatures.  I don't necessarily want to call out any companies on this
list, but there are at least 3 large ESPs that in aggregate send billions
of messages per day that are DKIM signed with rsa-sha1 (and only rsa-sha1).

Getting this small number of ESPs to change to rsa-sha256 will fix this
problem for email sent from a very large number of domains.  On the
referenced corpus I'm pretty sure it would address both the d=github.com
and d=travis-ci.org signed messages.  And I suspect it would represent a
much larger fraction of rsa-sha1 signed email in the average email user's
inbox.

One of the valuable services that DCRUP can provide is demonstrating to
these ESPs that SHA-1 support is actively being deprecated, and that they
will need to make the shift to rsa-sha256 to ensure future deliverability.
It's M3AAWG this week, and I'm planning on having that conversation with
several folks from companies that are using rsa-sha1, to encourage them to
make the change sooner rather than later.

Similarly, I'd suggest that participants in mailing lists that are signing
with rsa-sha1 reach out to the mailing list administrators and encourage
them to make this change.

Best,

Peter

On Mon, Jun 12, 2017 at 5:10 PM, James Cloos <cloos@jhcloos.com> wrote:

> >>>>> "BL" == Brandon Long <blong@google.com> writes:
>
> BL> It would be good to know where the source is
>
> When calculating those, I grep(1)ed for /^DKIM-Signature:/.
>
> 195704 sha1 dkims which had their d= on that line.  The result of:
>
> :; grep sha1 dkim-lines|tr \  \\n|grep ^d=|sort|uniq -c|sort -nr|head
>
> is:
>
>   55567 d=gcc.gnu.org;
>   46147 d=github.com;
>   32100 d=sendgrid.me;
>   28126 d=sourceware.org;
>    6865 d=
>    4560 d=listbox.com;
>    1980 d=pobox.com;
>     922 d=zx2c4.com;
>     891 d=itsqueeze.com;
>     661 d=travis-ci.org;
>
> The gcc.gnu.org ones are from the mailing lists @gcc.gnu.org, including
> the automated ones like gcc-testresults and gcc-bugs.
>
> For github, roughly half were a=rsa-sha1 and half a=rsa-sha256.
> I found a sha2 example from 2017/02/14 and a sha1 from 2017/02/15.
> The sha1 included a Received line referencing sendgrid between the
> ones referencing github itself.
>
> Getting those top four fixed would reduce significantly the sha1 cases.
>
> In all, there were 4217 distinct d= lines in the a=rsa-sha1 subset.
>
> But, again, the counts ignored cases where the a= and/or d= were not on
> the first line of the DKIM-Signature: header.
>
> -JimC
> --
> James Cloos <cloos@jhcloos.com>         OpenPGP: 0x997A9F17ED7DAEA6
>
> _______________________________________________
> Dcrup mailing list
> Dcrup@ietf.org
> https://www.ietf.org/mailman/listinfo/dcrup
>

-- 

[image: logo for sig file.png]

Bringing Trust to Email

Peter Goldstein | CTO & Co-Founder

peter@valimail.com
+1.415.793.5783 <(415)%20793-5783>