Skip to content

Est. 2011ยทMicrosoft Partner ID 7033487ยทDelivery under 3 minยทSupport 7 days a week

Your vault is empty.

License Error 550 5.7.64

550 5.7.64 TenantAttribution: Exchange Online refuses relay from your server

13 min read Updated October 5, 2026 Exchange Server

Fix it now

550 5.7.64 is published as TenantAttribution, Relay Access Denied. Exchange Online could not match the connecting server to an inbound connector in your tenant, so it treated a hybrid partner as an anonymous internet host. In almost every case the certificate your server presents no longer matches the value on the connector.

First line in Exchange Online PowerShell, second on the on-premises server, third in Exchange Online once you know the right name

Get-InboundConnector | Format-List Name,Enabled,ConnectorType,TlsSenderCertificateName,SenderIPAddresses
Get-ExchangeCertificate | Format-List Thumbprint,Subject,CertificateDomains,Services,NotAfter
Set-InboundConnector -Identity "Inbound from Contoso" -TlsSenderCertificateName mail.contoso.com
  1. Compare the two. The name in TlsSenderCertificateName has to appear as the subject or as one of the certificate domains on the certificate the server presents for SMTP.
  2. If you are not certain which certificate the server actually presents, read the send connector protocol log. (Get-SendConnector "Outbound to Microsoft 365").SourceTransportServers | foreach {get-transportservice $_.name} | Select-Object name,SendProtocolLogPath tells you where it is, and the log records the certificate subject and thumbprint.
  3. Realign the tenant side with Set-InboundConnector, using the name the certificate genuinely carries.
  4. Or rerun the Hybrid Configuration Wizard and select the new certificate on the transport certificate page, which sets the value for you.
  5. Wait a few minutes, send a fresh test message and read the new response rather than a queued one.

If the response mentions ATTR36 as well, the certificate name is not just mismatched: it is not a domain validated in your tenant. Add that domain in Microsoft 365, or reissue the certificate with a name that matches an accepted domain.

If a test message to an external recipient is delivered, you are done. If it still bounces, the next section explains what attribution is doing and where else it can fail.

Why it happens

When your server relays outbound through Microsoft 365, it is asking a public service to accept mail from an internet host and then treat it as internal. Anonymous senders do not get that. Exchange Online has to identify the connecting server as belonging to a specific tenant first, and it does so during the TLS handshake rather than afterwards.

The mechanism is the certificate. Your server presents one, Exchange Online reads its subject and alternative names, and looks for an inbound connector whose TlsSenderCertificateName matches. Microsoft’s published cause for this NDR is exactly that going wrong: an inbound connector is being used to receive messages from your on-premises environment, and something has changed on premises that makes the connector’s configuration incorrect – the host name in the certificate no longer matching the source server, or the source IP no longer matching the addresses on the connector.

When nothing matches, the session is not dropped. It continues as ordinary anonymous internet mail, which is adequate for messages addressed to your own accepted domains and useless for anything else. As soon as the message needs relaying onward, the service refuses. TenantAttribution names the stage that failed, not the recipient or the content.

There is a second documented variant worth knowing, because its fix is different. Where the certificate’s subject is not a domain validated in your Microsoft 365 tenant at all – CN=contoso.com when contoso.com has never been added, or CN=hostname.contoso.com when only contoso.com is registered – the response carries ATTR36. Microsoft’s preferred fix there is to add the domain in Microsoft 365 under Settings then Domains, and the alternative is to have the certificate reissued with a name that matches an accepted domain.

The certificate was renewed and the subject no longer matches

You have this one if Mail flow was healthy until a renewal date, and CertificateDomains on the new certificate does not contain the string held in TlsSenderCertificateName.

  1. Read the connector value: Get-InboundConnector | Format-List Name,TlsSenderCertificateName.
  2. Read the new certificate: Get-ExchangeCertificate -Thumbprint <thumbprint> | Format-List Subject,CertificateDomains.
  3. Update the connector to a name the certificate genuinely carries: Set-InboundConnector -Identity "<name>" -TlsSenderCertificateName mail.contoso.com.
  4. Confirm the certificate is enabled for SMTP on premises: Enable-ExchangeCertificate -Thumbprint <thumbprint> -Services SMTP.

A wildcard certificate works, but the connector value has to be written in the form the certificate actually presents. Read CertificateDomains rather than assuming.

The certificate name is not a validated domain in the tenant

You have this one if The response carries ATTR36 as well as 5.7.64, and the certificate subject names a domain or host that is not in your tenant.

  1. Preferred: sign in to Microsoft 365 with administrator permissions, go to Settings then Domains, and add the domain used on the certificate.
  2. Where the subject is hostname.subdomain.contoso.com, add subdomain.contoso.com rather than the full host name.
  3. Alternative: have the certificate reissued with a name that matches an accepted domain in the tenant, install it on the Exchange server used for hybrid mail flow, and confirm the FQDN on the connector.
  4. Rerun the Hybrid Configuration Wizard afterwards so the Exchange Online connector is set correctly.

The send connector offers a different certificate than you think

You have this one if Connector and certificate both look right, and the server has several certificates enabled for SMTP.

  1. Find the protocol log path: (Get-SendConnector "Outbound to Microsoft 365").SourceTransportServers | foreach {get-transportservice $_.name} | Select-Object name,SendProtocolLogPath.
  2. Read that log for a recent session and note the certificate subject and thumbprint actually sent.
  3. Pin the certificate on the send connector with its TlsCertificateName property, which takes the issuer and subject in the combined form Exchange expects.
  4. Restart the transport service so the selection takes effect, then check the log again.

A gateway or firewall re-terminates the TLS session

You have this one if The certificate on the Exchange server is correct, and the tenant reports a certificate name nobody recognises, or none at all.

  1. Trace the outbound path. Any security appliance, cloud filter or inspecting firewall in the way presents its own certificate to Microsoft 365.
  2. Either pass the session through without re-terminating it, or set the inbound connector to the appliance’s certificate name.
  3. If the appliance cannot present a usable certificate, populate SenderIPAddresses on the connector and rely on address-based attribution instead.
  4. Retest from a server that uses the same path, not from one that bypasses it.

The connector was rebuilt, renamed or disabled

You have this one if Get-InboundConnector shows no OnPremises connector, one with Enabled set to false, or two competing after a wizard rerun.

  1. Remove duplicate or orphaned OnPremises connectors left by earlier runs, one at a time.
  2. Rerun the Hybrid Configuration Wizard from a current on-premises server so both ends are written together.
  3. If you would rather not run the wizard, enable the surviving connector and set TlsSenderCertificateName by hand.
  4. Check RestrictDomainsToCertificate and SenderDomains; a narrow restriction rejects sessions that would otherwise attribute correctly.

The wizard changes both the tenant and the on-premises organisation. Record your current connector settings first so you can see afterwards what it changed.

Full reference

Reading the symptom to find the broken side

What you observe Where the fault sits
Mail to your tenant domains flows, external relay fails Attribution failed and the session was downgraded to anonymous
Everything fails, including mail to tenant domains The connector is disabled or deleted, or TLS is failing before attribution
It broke on the day a certificate was renewed The subject or alternative names changed and the connector was not updated
The response mentions ATTR36 The certificate name is not a validated domain in the tenant
Free/busy fails as well A different problem. See the organisation relationship section below

The codes that travel with this one

Code What is published
550 5.7.64 TenantAttribution; Relay Access Denied. An inbound connector whose configuration no longer matches the source server’s certificate or IP address
550 5.7.57 Client not authenticated to send anonymous mail during MAIL FROM. A submission to smtp.office365.com must authenticate as a user with an Exchange Online mailbox
1007 AccessDenied A Hybrid Configuration Wizard failure. Microsoft’s published cause is local system time five minutes or more out of sync with a valid time server
ErrorNoFreeBusyAccess An EWS response with the text “The caller does not have access to free/busy data”. The requesting user’s SMTP domain is not in the organisation relationship’s DomainNames
550 5.7.51 Nothing published
550 5.7.54 Nothing published

1007 AccessDenied is the one people misfile. It is not a mail flow error and it is not about permissions between the two organisations: Microsoft documents it as the wizard being unable to federate because the local system time is five minutes or more out of sync with the federation servers. The fix is w32tm /resync and a properly configured time source, not a connector change.

When free/busy fails as well

ErrorNoFreeBusyAccess looks related because it appears in the same hybrid environments, and it is a separate fault with a separate fix. Microsoft’s article on it quotes the EWS response – a MessageText of “The caller does not have access to free/busy data” with ResponseCode ErrorNoFreeBusyAccess – and gives the cause as the SMTP domain of the user requesting availability not being present in the organisation relationship’s DomainNames. The documented fix adds it:

$OrgRel = Get-OrganizationRelationship Contoso
$OrgRel.DomainNames += "contoso.com"
Set-OrganizationRelationship $OrgRel.Name -DomainName $OrgRel.DomainNames

The common version of this is a request that uses the wrong suffix – a user asking as someone@contoso.com when the relationship only knows contoso.mail.onmicrosoft.com, or the reverse. Check which domain the requesting mailbox’s primary SMTP address actually uses before adding anything.

Two codes with nothing published

550 5.7.51 and 550 5.7.54 both turn up in this area and neither appears in Microsoft’s Exchange Online NDR reference or anywhere else in its documentation. 5.7.54 in particular is commonly seen and commonly explained on the internet; none of those explanations comes from the vendor. Record what your own NDR text says alongside the code, because the text is the part that is actually authoritative, and diagnose from the connector configuration rather than from the number.

Confirming a change took effect

  • Give a connector change a few minutes and then send a new message. Do not judge it from a message that was already queued, because that may still carry the earlier decision.
  • Use Get-MessageTraceV2 in Exchange Online to see how the new message was handled. Microsoft’s own reference states that Get-MessageTrace is replaced by Get-MessageTraceV2 and will eventually be deprecated.
  • Check the message was accepted through the inbound connector you changed, not through a different one you had forgotten about.
  • Re-read Get-InboundConnector | Format-List TlsSenderCertificateName and confirm the value appears in CertificateDomains on the live certificate rather than on the one you meant to install.
  • Run a free/busy lookup across the boundary as well, so you know whether you have one problem or two.

When a licence is the actual fix

Repairing this connector costs nothing, and that should be said plainly before anything else: reading two values and setting one is a five-minute job and it keeps your current design intact. The only unavoidable cost is a replacement public certificate if yours has expired, and that comes from a certificate authority rather than from a licence reseller. There is a longer-term question underneath it, though. If the on-premises server exists mainly to hold the last few mailboxes and relay for applications, this plumbing is a cost you pay again at every certificate renewal, and every renewal is a chance to break outbound mail for a morning. Moving the remaining mailboxes to Microsoft 365 removes the attribution problem entirely, because there is no longer an on-premises server to identify. Microsoft 365 Business Standard is the usual choice at that size. Arco can supply it and check which plan covers the mailbox features you rely on today, but this is a decision to take on its own merits rather than in the middle of an outage.

Every code this article covers

Code What it points at Source
550 5.7.64 TenantAttribution; Relay Access Denied. The inbound connector’s configuration no longer matches the source server: the certificate host name, or the source IP address Microsoft Learn
1007 AccessDenied A Hybrid Configuration Wizard failure provisioning Exchange to the Partner STS. Microsoft’s published cause is local system time five minutes or more out of sync with a valid time server Microsoft Learn
ErrorNoFreeBusyAccess An EWS availability response, “The caller does not have access to free/busy data”. The requesting user’s SMTP domain is not in the organisation relationship’s DomainNames Microsoft Learn
550 5.7.57 Client not authenticated to send anonymous mail during MAIL FROM. Submissions to smtp.office365.com must authenticate as a user with an Exchange Online mailbox Microsoft Learn
550 5.7.51 Seen on submissions refused by Exchange Online. Microsoft publishes no meaning for this enhanced status code; read the NDR text that accompanies it not published by the vendor
550 5.7.54 Seen on relay refusals. Microsoft publishes no meaning for this enhanced status code, despite how often it is quoted; read the NDR text not published by the vendor

Confirm the fix worked

  1. A test message from an on-premises mailbox to an external recipient is delivered rather than returned.
  2. Get-MessageTraceV2 in Exchange Online shows that message accepted through the inbound connector you changed.
  3. Get-InboundConnector | Format-List TlsSenderCertificateName returns a value that appears in CertificateDomains on the certificate the server actually presents.
  4. The send connector protocol log shows the certificate you intended being offered.
  5. A free/busy lookup across the hybrid boundary succeeds, which confirms the trust as well as the mail flow.

Questions people ask about this

How long does a connector change take to apply?

Give it a few minutes and retest with a new message. Do not judge the result from a message that was already queued, because that may still be carrying the decision made when it was submitted.

Must I re-run the Hybrid Configuration Wizard at every certificate renewal?

No. If the new certificate carries the same subject and alternative names, nothing needs to change. The wizard is for when names change, or when several settings have drifted at once – and Microsoft’s own guidance is to rerun it and select the new certificate on the transport certificate page when a name has changed.

Does fixing this cost anything?

The connector fix is free. The only cost that can appear is a replacement public certificate if yours expired, and that comes from a certificate authority. A Microsoft 365 licence is relevant only if you decide to retire the on-premises server instead.

Can I use address-based attribution instead of a certificate?

You can populate SenderIPAddresses on the inbound connector, and it is the practical answer when an appliance re-terminates TLS. It is weaker than certificate matching, so keep the list tight and review it whenever your addressing changes.

What is 1007 AccessDenied doing in this article?

It appears in the same hybrid logs and is routinely misread as a permissions problem between the organisations. Microsoft documents it as the wizard failing to federate because the local system time is five minutes or more out of sync with the federation servers, so the fix is the clock.

Related error codes

Was this article helpful?

Your feedback helps us improve our documentation.

Related articles

Free Fix TargetUserAlreadyHasPrimaryMailboxException and other stuck move requests Free Fix Error -1206 JET_errDatabaseCorrupted: a non-database file or a corrupt database Free Fix Event ID 7009: Exchange setup times out waiting for a service to start License Error Error -1018 JET_errReadVerifyFailure: page checksum damage in a mailbox database
โ† Back to Knowledge Base