The servicedesk got complaints that email was not delivered to an organization, and that an NDR was never generated. The sending user never knew this, until the sender and receiver talked on the phone. Our organization was Exchange 2016 on-premises with a Cisco IronPort as a gateway to the Internet. The other (receiving) organization was in Exchange Online.
The receiving organization was in Exchange Online, and had already enabled DANE for inbound email messages (see my previous blog post on this). I checked multiple organizations in Exchange Online that have DANE enabled (including hotmail.nl) and they all failed.
One organization has a Fortra Clearswift in front of their environment that has DANE enabled, and we were able to send email to this particular domain.
And to make it more complex, other organizations with an IronPort gateway were able to successfully send email messages to these domains.
At this point still no clue whether it is a Microsoft issue, or an IronPort issue or something specific to our organization.
When checking the DANE SMTP service for the domains involved everything looks fine as shown in the following screenshot:
When checking the IronPort logs, the following cryptic and non-explaining error was logged:
MID 1614647 (DCID 600113) DANE failed for Hotmail.nl. Reason: 4.0 - Other network problem.
Also shown in the following screenshot:
The same error was logged for the other DANE enabled domains as well.
So, DANE fails on the IronPort, but the tools to check DANE all reported DANE was good. Also, I was able to send mail to these domains using Gmail. I always say when it works in Gmail, everything is ok.
When checking the DANE configuration on the command prompt on the IronPort it looks more like a DNS issue as shown in the following screenshot:
But when checking the DNS record (_25._tcp.exchangelabs-nl.y-v1.mx.microsoft) with MXToolbox, everything is green again as shown in the following screenshot:
After checking with the network department it turned out that there was an IPS solution implemented and the network engineer knew about an old CVE, dating back to 2013 (CVE-2013-4466) that warns for a situation where a buffer flow can occur in the dane_query_tlsa function when more that four DANE entries are returned.
The CVE is 11 years old, but the IPS still had this implemented, and when more than four entries were returned, everything was discarded and the email was lost. Removing this from the IPS and everything works fine.
Note. The RFC for DANE does not mention the amount of entries that can be returned, so more than four should not be a problem.
Outbound DANE in Exchange Online is available for some time now and I wrote about this before in the following blogpost: DNSSEC and DANE support in Exchange server and Exchange Online.
Microsoft recently announced the public preview (!) of inbound SMTP DANE for Exchange Online and it’s fairly easy to configure.
Implementing DANE consist of two steps:
Enabling DNSSEC in Exchange Online.
Enabling DANE in Exchange Online.
I will show both in the following sections.
Enable DNSSEC in Exchange Online
An important prerequisite of course is that you have DNSSEC up and running. If you have, you can open an PowerShell window and connect to Exchange Online.
The output of the command will show the (new) MX record for the domain as shown in the following screenshot:
Add the new MX record to your domain, but give it a lower priority than the existing MX record (which typically should also point to Exchange Online).
Important: If you have configured MTA-STS for your inbound mail in Exchange Online, DO NOT FORGET to change the MTA-STS policy to reflect the new MX record!
My previous blogpost was about DANE and discussed how DANE can be used to make TLS negotiations between mailservers more secure. Another topic in this area is MTA-STS. MTA-STS stands for Message Transfer Server – Strict Transport Security and MTA-STS is a mechanism to enforce the use of TLS and the use of a valid 3rd party server certificate.
MTA-STS, DNS and Policies
When an MTA-STS capable servers wants to send an email, it first retrieves the MX record for the recipient domain. The second step is that the sending server check for an MTA-STS record in DNS. This record looks like:
_mta-sts.exchangelabs.nl. IN TXT "v=STSv1; id=202306242147;"
The id is an identifier and defines the version of the MTA-STS record when changes are made to the MTA-STS record. A good practice is to create an identifier based on the date and time of the last change. In this example, it is June 24, 2023 at 9:47pm.
The next step is that the sending servers looks for a policy. This policy is not stored in DNS, but on a website. The URL for this policy looks like: https://mta-sts.exchangelabs.nl/.well-known/mta-sts.txt. The subdomain mta-sts, the filename mta-sts.txt and the directory .well-known (including the dot) directory are mandatory for the MTA-STS policy. It must also be secured using a valid 3rd party server certificate.
Note. If you have configured DANE for inbound email in Exchange Online, your MX record should be something like Exchangelabs-nl.y-v1.mx.microsoft.
The MTA-STS policy is structured as follows:
Version identifies the version of MTA-STS but must always be STSv1 (for now at least).
Mode defines how the policy must be applied:
Enforce: Sending MTAs MUST NOT deliver the message to hosts that fail MX matching or certificate validation or that do not support STARTTLS.
Testing: Sending MTAs that also implement the TLSRPT (TLS Reporting) specification [RFC8460] send a report indicating policy application failures (as long as TLSRPT is also implemented by the recipient domain); in any case, messages maybe delivered as though there were no MTA-STS validation failure.
None: In this mode, Sending MTAs should treat the Policy Domain as though it does not have any active policy.
MX defines all MX records in use by the recipient domain. This can be one entry, but it can hold multiple MX records, each on a separate line as shown in the policy above.
Max_age defines the time (in seconds) that the MTA-STS policy can be cached by a mail server. In this example, the policy is cached for 604800 seconds, which equals to 1 week. When a sending server must send a new email within a week, the policy is still cached. After checking the MX record the server retrieves the TXT record from DNS (as explained in the second step above) and when the identifier has not changed it uses the policy that is cached. If the identifies has changed within the lifetime of the cached policy, a new policy is downloaded.
So, in my example an MTA-STS capable mail server will check the MTA-STS policy and only connects to my mail server using TLS 1.2 (this is enforced with MTA-STS when mode is set to ‘enforce’) and only when a certificate that matches the FQDN is used. When authentication fails for an entry, the sending server continues with the next line in the policy, in my example with the MX record pointing to Exchange Online.
An interesting option in MTA-STS is reporting. DMARC has a reporting function as well, but reports are only sent by receiving domains. Reporting in MTA-STS is performed daily by sending mail servers that supports MTA-STS and TLS-RPT.
To configure the reporting functionality, create a mailbox in Exchange 2019 or Exchange Online and assign it an email address like TLSReports@Exchangelabs.nl. The next step is to configure the following DNS TXT record:
_smtp._tls.exchangelabs.nl. 3600 IN TXT v=TLSRPTv1;rua=mailto:TLSReports@exchangelabs.nl
There are several online tools available for checking the MTS-STS record. Just like in my other blogs, I often use MXToolbox to check for DNS records as shown in the following screenshot:
The first one is obvious, the second line is a Microsoft IP address, the third line is a Yahoo IP address. So, Google, Microsoft and Yahoo are using MTA-STS when sending email.
MTA-STS versus DANE
MTA-STS and DANE share a common concept, that is to secure the (initial) communication between mail servers. The ‘problem’ with DANE is that is relies on DNSSEC and the global roll-out of DNSSEC is very slow (to put it mildly).
MTA-STS was developed to overcome the slow roll-out of DNSSEC (since it does not use DNSSEC of course). MTA-STS can be seen as a ‘light-weight’ version of DANE and it will be sufficient for most customers.
And how about Exchange?
Just like with DANE, the ugly part is that Exchange 2019 does not support MTA-STS. You can configure the MTA-STS record in DNS and the policy on a website so that MTA-STS capable servers use your Exchange 2019 server safely. But for sent messages by Exchange 2019, MTA-STS is a no-go, it does not support it and most likely will never do.
Exchange Online on the other hand does support MTA-STS (since the beginning of 2022) for both inbound and outbound messages. The only thing you must do to enable it for inbound messages is create the TXT record in DNS and create and publish the MTA-STS policy.
In the past I have written about SPF, DKIM and DMARC for email authentication in Exchange server to improve e-mail security. Additional topics to improve security are secure DNS (DNSSEC) and DNS-based Authentication of Named Entities (DANE). Where SPF, DKIM and DMARC are focusing more on the email messages and the sending hosts they are coming from, is DANE more focusing on setting up the TLS connection between mail servers. DANE has a dependency on DNSSEC but since my focus is on Exchange and Exchange Online, I don’t discuss DNSSEC and leave that to my provider (DANE only works when DNSSEC is enabled).
When two mailservers setup a connection, they negotiate a common TLS protocol and this handshake as it is called is unencrypted. As such, this handshake is vulnerable for man-in-the-middle or downgrade attacks.
DANE is using a special DNS record (a TLSA record, Transport Layer Security Authentication) where an organization can publish information regarding the certificate on the mail server and thus the TLS options available.
The TLSA record for my own domain is shown in the following example:
_25._tcp.smtphost.exchangelabs.nl IN TLSA 3 1 1 50167c478a2f536a88ee9ec232b14b0d223c2d5bdc837451eee104c153376cbe
When a mailserver wants to send email to my Exchange environment it retrieves the MX record first which is smtphost.exchangelabs.nl (this is an Exchange 2019 Edge transport server, with a Digicert certificate that matches the servername). When DNSSEC is not used at the recipient domain, opportunistic TLS is used. When DNSSEC is used at the recipient domain, a DANE capable mail server will check for a TLSA record.
When a TLSA record is found and retrieved, the sending server sets up a TLS connection with the receiving server, which in turn returns the fingerprint of the certificate. The sending server compares the fingerprint with the information found in DNS and when it matches, the connection is established and communication between the two servers continue, and the email is sent.
So, how do you create the TLSA record? There are publicly available webservers that can create TLSA records. Shumon Huque is a software engineer and technologist based in Washington DC who has a site that can generate TLSA records. As a bonus his site can also check TLSA records.
To generate a new TLSA record, navigate to his website on https://www.huque.com/bin/gen_tlsa and fill out the necessary information as shown in the following screenshot:
The easiest way to create a PEM format X.509 certificate is using the MMC certificate snap-in. Select your certificate and export it. Select the ‘No, do not export the private key’ option, select the ‘Base-64 encoded X.509 (CER)’ option and export the certificate. You can use Notepad to open the certificate export, and copy-and-past this into the PEM format X.509 textbox.
Select the port number (25), the protocol (TCP, not SMTP) and enter the domain name. This is the FQDN of the receiving server, so in my example it is smtphost.exchangelabs.nl and click Generate.
It will show the generated TLSA record, including the certificate information as shown in the following screenshot:
For my domain the following TLSA record is generated:
_25._tcp.smtphost.exchangelabs.nl. IN TLSA 3 1 1 a63d74fc7ec1acad702017d13479a1b60b36f234ccb96b90f97c9619ba2c91ab
How is this TLSA record structured?
TCP port 25 and the FQDN of the server make up the first part of this TLSA record.
The first number after the TLSA text is the certificate verification or the certificate usage on the server and this number can have 4 values:
0 – CA Constraint. The certificate provided when establishing the TLS connection must be issued by the listed root-CA or one of its intermediate Certificate Authorities.
1 – Service Certificate constraint. – The certificate used must match the TLSA record, and it must also pass the certification path validation to a trusted root-CA.
2 – Trust Anchor Assertion. The TLSA record must match the certificate of the root CA, or one of the intermediate CAs of the certificate in use by the mailserver.
3 – Domain Issued Certificate. The TLSA record matches the used certificate itself. The certificate does not need to be signed by other parties and as such a self-signed certificate can be used (this is interesting on an Exchange server).
In my TLSA record a value of 3 is used for the certificate usage, so the TLSA record must match the certificate on the Exchange 2019 server.
The second number that is used is the selector. The selector can have two values:
0 – The entire certificate is used for matching.
1 – Only the public key of the certificate is used for matching.
In my TLSA record the value is 1 for the selector, so only the public key is used.
The third number that is used is the matching type. The matching type can have three values:
0 – the entire information select is present in the certificate associated data (the last text string in the TLSA record)
1 – The SHA-256 hash of the public key of the certificate must match the certificate associated data.
2 – The SHA-512 hash of the public key of the certificate must match the certificate associated data.
In my TLSA record a value of 1 is used for matching, to a SHA-256 has of the public key of the Exchange server certificate must match the data in the TLSA record.
The next step is to add the generated TLSA record in DNS and check the TLSA record once added.
You can use the same site (https://www.huque.com/bin/danecheck-smtp) to check the TLSA record. The site queries the MX record for your domain and checks the accompanying TLSA record. The result is shown in the following screenshot:
In this blog I am discussing DANE for mail server, but it can also be used for other services. For a website a TLSA record can be:
_443._tcp.www.exchangelabs.nl. IN TLSA 3 1 1 a63d74fc7ec1acad702017d13479a1b60b36f234ccb96b90f97c9619ba2c91ab
Or for IMAP it can be:
_995._tcp.mail.exchangelabs.nl. IN TLSA 3 1 1 a63d74fc7ec1acad702017d13479a1b60b36f234ccb96b90f97c9619ba2c91ab
The ugly part is that Exchange 2019 does not support DANE out of the box. For incoming mail, you can configure the TLSA record based on the certificate that is installed on the Exchange (Edge Transport) server. A DANE capable mail server automatically checks for a TLSA record and when available, use this to setup the TLS connection.
Exchange Online on the other hand only supports DANE for outbound connections at the time of writing. As an Exchange Online customer, there’s nothing you must configure, it’s available for everyone. DANE for inbound mail will become available in the future, but when is unclear. According to Microsoft it is ‘near future’ but unfortunately this is already the case for quite some time.
One final (and very important) closing note: When you have configured a TLSA record for your (inbound) Exchange server you MUST update your TLSA record when renewing the Exchange certificate. You must also do this when the Exchange server is used in a hybrid configuration, since Exchange Online will check for a TLSA record. Mail flow from Exchange Online to Exchange 2019 will halt and the following error message is shown in Exchange Online message trace:
Reason: [{LED=450 4.7.323 tlsa-invalid: The domain failed DANE validation [Message=450 4.7.323 tlsa-invalid: The domain failed DANE validation] [LastAttemptedServerName=smtphost.exchangelabs.nl] [LastAttemptedIP=185.116.41.45:25] [SmtpSecurity=11;-1] [HE1EUR01FT069.eop-EUR01.prod.protection.outlook.com 2023-06-06T10:23:06.0. OutboundProxyTargetIP: 185.116.41.45. OutboundProxyTargetHostName: smtphost.exchangelabs.nl
This is solved when a new TLSA record is generated and published in DNS.
Jaap Wesselius 
You must be logged in to post a comment.