6
The Microsoft Problem
Problem reported by Douglas Foster - 3/26/2024 at 1:48 PM
Submitted
Here is a list of the ways that Microsoft practices are complicating my efforts to disposition messages correctly.  It seems like others should be aware of the problems as well.

  1. Onmicrosoft.* parent domains are not listed in the Public Suffix List used for DMARC.  This can allow one Office365 client to impersonate another, if a DMARC policy is published with relaxed alignment.  Fortunately, SmarterTools has corrected this omission in the list that they embed within SmarterMail.
     
  2. Microsoft ARC Sets cannot be trusted indicators of message verification status.  They sometimes report false PASS, and sometimes false FAIL.    In one recent example, the message actually failed SPF, had no signature, and no DMARC policy, but Microsoft asserted pass for SPF, DKIM, and DMARC.
     
  3. Onmicrosoft.com domains, using Microsoft servers, are regularly used to send spam.  Often the From domain name is obviously computer generated.    Microsoft seems to have no controls in place to prevent this abuse of their servers.
     
  4. Outlook.com DMARC reports are inaccurate, because they send reports based on the message state AFTER going through the client’s spam filter (which breaks SPF).  If the spam filter also adds annotations, the process also breaks DKIM.   If the DMARC policy is “reject”, their report says that the message was rejected.  In my experience, they do not act on the reject result that they report, since the client is configured with a non-Microsoft spam filter.   But getting a “reject” result from an important business partner set off alarms when I saw it.
     
  5. Blind Carbon Copies are supposed to be truly blind, in both the user-visible and internal versions of a message.   To ensure this, organizations are expected to remove the bcc information to prevent unauthorized information disclosure.  (Since BCC should be removed, BCC is not allowed in a DKIM signature’s header list.)   However, Microsoft copies the BCC information into a custom header named X-MS-Exchange-CrossPremises-BCC.  This causes leakage of information that should not be leaked, albeit only in the internal version of the message.   I have contacted two Office365 client organizations to make them aware of the problem, but have received no follow-up from them.
  6. Some Microsoft servers submit a message and then disconnect without waiting for an SMTP response code, but assume that the message was accepted.  Standards-compliant servers will treat the disconnect as a network glitch, discard the message, and wait for the transmission to be repeated.   As a result, the message is lost.  The workaround is for the receiving server to process the message as if the SMTP result code had been sent.   For situations where the problem is a true network glitch, a standards-compliant sender will retransmit the message, and because of the Microsoft workaround, the recipient will receive a duplicated message.
  7. Cell-phone versions of Outlook always connect to a Microsoft server instead of connecting to the clients' autodiscover server.   This routing is not explained to the application user.   This pass-through process means that both credentials and message content are presumably leaked to Microsoft.

10 Replies

Reply to Thread
0
John C. Reid Replied
I don't know if you have seen this or not Douglas, but we have also observed more than one of their IP addresses on RBL lists. I can't remember which one off the top of my head, but it was a large list like Spamcop. We have seen this recently and multiple times.

Of course end users just can't understand why they don't get the email from the sender using MS365. I mean Microsoft can't be doing anything wrong, so it must me my provider doing something wrong in not delivering the message, right? (I would have tagged the previous sentence sarcasm, except that is actually, exactly what the perception from end users tends to be.)
John C. Reid / Technology Director John@prime42.net / (530) 691-0042 1300 West Street, Suite 206, Redding, CA 96001
1
Linda Pagillo Replied
Hey guys! 

I have been seeing a lot of what John is seeing. Spam coming from MS IPs.

Also Doug, I have reached out to our Microsoft specialists to see if I can get any further info about these things.
Linda Pagillo Mail's Best Friend Email: linda.pagillo@mailsbestfriend.com Web: www.mailsbestfriend.com Authorized SmarterTools Reseller Authorized Message Sniffer Reseller
0
Ron Raley Replied
Same here as John.  Several times last month Office 365 and Hotmail (outbound.protection.outlook.com) showed up on SpamCop, SORBS, and Backscatter.

Trying to tell the customer that their IP is sending spam can be cumbersome.  The mindset is that they are using "Microsoft" so they must not be the problem.  The problem resides with us, a little Mom & Pop shop using SmarterMail.  We ask them to start a ticket with Microsoft and they become angered.
1
Douglas Foster Replied
All of the spam coming from computer-generated *.onmicrosoft.com domains should put their addresses on the RBL lists.  I block everything from *.onmicrosoft.com except for two domains that are known business partners.
2
AWRData Replied
It seems to me there should be no emails coming from onmicrosoft.com domains, as these are tenant internal AD aliases.  Therefore, I see no problem blocking them.
2
Linda Pagillo Replied
AWR.. I have seen plenty MS customers actually use the onmicrosoft.com domain as their main email domain. This is why I'm not blocking them. Should they be using it? Not really. But they do :(
Linda Pagillo Mail's Best Friend Email: linda.pagillo@mailsbestfriend.com Web: www.mailsbestfriend.com Authorized SmarterTools Reseller Authorized Message Sniffer Reseller
1
@Douglas, I would consider sending that list, especially #5, over to "Bleeping Computer" and have them write up a little article about it. THAT might get MS attention. 

www.HawaiianHope.org - Providing technology services to non profit organizations, low income families, homeless shelters, clean and sober houses and prisoner reentry programs. Since 2015, We have refurbished over 11,000 Computers !
1
Douglas Foster Replied
Update Weds April 3, 2024
Added bullet points 6 and 7 to the list, for completeness, even though they have been discussed previously on this forum:
#6: Microsoft servers do not always wait for SMTP result code when submitting a message
#7: Some Outlook implementations leak credentials and data to Microsoft.
0
John C. Reid Replied
Microsoft announced at least a couple years ago that #7 would be for all clients in the future. I think this may have already happened. There was a blog post here on the SmarterTools website about it. I have seen some behavior from a few of my clients that use the MS365 version of Outlook that would suggest it is already doing a proxy through Microsoft servers rather than directly connect to the mail server. In fact is was just a few months back I noticed Microsoft owned IPs connecting via POP3 and IMAP to my server.
John C. Reid / Technology Director John@prime42.net / (530) 691-0042 1300 West Street, Suite 206, Redding, CA 96001
1
AWRData Replied
John is correct.  While I have not seen these blog posts, I can confirm that over the past couple of months, all autodiscover traffic (mostly EAS) I have been troubleshooting for customers, and my family's Windows Mail/New Outlook IMAP traffic has been coming from Microsoft.  I have not yet seen this with Outlook proper (2016-2021,) but I suspect it is coming.

I am encouraging my customers' IT to look into K9 Mail for Android and alternative IMAP clients for Windows, such as Thunderbird.

It has been 20 years since the Microsoft Love In of the 2000s, with which came free copies of dotNet development kits, Windows XP, Office 2003, and Server 2003, free TS2 seminars, TechNet subs and MAPS.  We all got on-board the Microsoft train and here we are: Governor Tarkin now runs the rails of technology and the grasp is tightening.

Reply to Thread