Is Google failing DKIM keys shorter than 512 bits?

Today’s Wednesday question comes from Andrew B. and got pushed to Thursday so I could check a few more facts.

Have @Gmail yet confirmed the @ReturnPath story that they’ll start failing weak DKIM sigs? RP cites no source: http://goo.gl/Rb5to  @hey4ndr3w

The answer is that no one from Gmail has publicly confirmed that they’re failing to authenticate mail signed with weak DKIM keys. But conversations with various Return Path folks indicate this will happen at some point.
As of the time of this writing, Google is currently marking DKIM mail signed with 512 bit keys with “dkim=pass” and delivering the mail to the inbox. This was confirmed on Twitter Wednesday afternoon.

In case you missed it earlier- there are blog posts circulating about Gmail failing 512 bit DKIM keys, but testing shows 512 bits still pass @evan_burke

I posted earlier about DKIM key lengths and how there are vulnerabilities with shorter signing keys. The most recent revision of the DKIM spec itself states that signers MUST use at least 1024 bits for signing when the keys are intended for long term use. Anyone who is signing with a key less than 1024 and who is not rotating keys regularly is technically not following the DKIM specification.
What does this really mean?
Right now, DKIM authentication lends legitimacy to the signing domain (the d= value). Some ISPs, like Gmail, are using this authentication to present information to the end user about whether or not an email is legitimate. ISPs can trust that the sender of DKIM signed mail is an authorized user of the domain in the d= field because only authorized users have access to the DKIM private key.
When the DKIM key is short (512 bits or less) it can be cracked with a little bit of processing power. The Wired article specifically states that DKIM keys of 512 can be cracked in about 3 days using $75 of Amazon cloud computing. That means that if AnEmailCompany is signing with a DKIM key of 512 bits then a malicious entity could crack that key and send DKIM signed mail spoofing AnEmailCompany. The malicious entity gets the benefit of AnEmailCompany’s domain reputation and AnEmailCompany may see a drop in inbox delivery as the malicious entity sends more spam.
In other words, while the DKIM signature may validate, the mail may not be signed by an authorized entity. It is reasonable for Google, or any other entity checking DKIM signatures, to evaluate the length of a key and the length of time it has been in circulation when deciding to believe authentication results.
In terms of delivery, I expect this means that Google will treat mail signed with insecure keys similarly to mail not signed with DKIM at all.  Delivery decisions (inbox? bulk folder?) will be based on other factors in the email. Positive domain reputation may not help inbox delivery of poor content. And, for some cases, this may result in “via” showing up in the message list for some email.
I don’t believe that “failing DKIM keys shorter than 1024″ means that they’re going to reject or bulk all mail signed with shorter keys. That’s going to hurt an awful lot of wanted mail, and like most mail providers Gmail wants to deliver mail their users want. But they may not trust mail signed with short keys as much as they’ll trust mail signed with secure keys.
===
Have a question you want me to answer? Tweet them to wise_laura or send them to Nov07@contact.wordtothewise.com

Related Posts

Setting up DNS for sending email

Email – and email filtering – makes a lot of use of DNS, and it’s fairly easy to miss something. Here are a few checklists to help:

Read More

Predictions for 2008

I did not have a lot of predictions for what will happen with email at the beginning of the year so I did not do a traditional beginning of the year post. Over the last 3 – 4 weeks, though, I have noticed some things that I think show where the industry is going.
Authentication. In January two announcements happened that lead me to believe most legitimate mail will be DK/DKIM signed by the end of the year. AOTA announced that approximately 50% of all email was currently authenticated. They did not separate out SPF/SenderID authentication from DK/DKIM authentication, but this still suggests email authentication is being widely adopted. AOL announced they will be checking DKIM on their inbound mail. I expect more and more email will be DKIM signed in response to this announcement.
Filtering. The end of 2007 marked a steady uptick in mail being filtered or blocked by recipient domains. I expect this trend to continue throughout 2008. Recipient domains are rolling out new technology to measure complaints, evaluate reputation and monitor unwanted email in ways that tease out the bad actors from the good. This means more bad and borderline email will be blocked. Over the short term, I expect to see more good email blocked, too, but expect this will resolve itself by Q2/Q3.
Sender Improvements. As the ISPs get better at filtering, I expect that many borderline senders will discover they cannot continue to have sloppy subscription practices and still get their mail delivered. Improved authentication and better filtering let ISPs pin-point blocks. Instead of having to block by IP or by domain, they can block only some mail from a domain, or only some mail from an IP. There are a number of senders who are sending mail that users do not want mixed with mail that recipients do want. Right now, if there is more mail that recipients want in that mix, then ISPs let the mail through. This will not continue to happen through 2008. Senders will need to send mail users actively want in order to see good delivery.
Less is more. A lot of other email bloggers have talked about this, and I will echo their predictions. Less email is more. Send relevant mail that your customers want. Target, target, target. Good mailers will not send offers to their entire database, instead they will send mail to a select portion of their database.
Feedback loops. Use of feedback loops by recipient domains will continue to grow.
Mobile email. More recipients will be receiving email on mobile devices.
Suggestions for 2008

Read More

Gmail and the via

I was hoping to have a detailed post up today about the conditions where gmail presents the user with a “via” but time seems to have gotten away from me. But I can give you the conclusions.

Read More