[ Forwarding & ARC ]

Email signature services and DKIM breakage

Separate visible signature insertion from DKIM signing, then check the integration path, MIME changes and domain alignment of the final message.

By the EmailAuth research team · 6 min read

[ Key points ]

  • A contact footer and a DKIM signature are different mechanisms.
  • Client-side and server-side insertion can sit at different points in the delivery path.
  • A provider signature can verify without aligning with the visible author domain.
  • Appearance, duplicate insertion and authentication require separate acceptance checks.
On this page

Separate the visible signature from the cryptographic one

An email signature service usually adds visible material such as a person's contact details, company logo or legal wording. DKIM adds a cryptographic header that binds selected message content to a signing domain. The shared word “signature” does not make these operations interchangeable. The key question is whether visible content is inserted before or after the DKIM signature that the receiver evaluates.

No signature-service account, connector or raw customer message was examined here. This article provides a diagnostic plan and a fictional ordering example. The original evidence gate remains open: Original and modified raw messages, signer placement and a before-and-after verification test. Collect these before attributing an incident to a service.

Locate where the visible material is inserted

A user's mail application can insert a signature during composition. A server-side service can instead receive a submitted message, add content and send it onward. Other workflows combine client insertion with server fallback. These are possible designs, not universal product behavior. Ask for the specific supported integration and the route actually used by the affected sender.

Microsoft's documentation distinguishes user-applied signatures from organization-wide mail-flow rules that add signatures, footers or headers. It demonstrates that insertion location can differ even within a familiar platform. It does not establish the relative order of an unknown third-party service and an organization's DKIM signer. Retain the provider's applicable integration guidance and verify the deployed sequence separately. Microsoft message additions

Read the signature that actually failed

Record the DKIM signing domain, selector, canonicalization and signed-header list for every signature present. A message may carry signatures from multiple processing stages, and each must be assessed independently. Do not summarize one failed original signature and one valid provider signature as a single unexplained “DKIM problem.” Preserve the result and domain associated with each check. RFC 6376 sections 3.5 and 4.2

For the body, compare the bytes after the selected canonicalization, not only the appearance of the email. Relaxed body canonicalization normalizes specified whitespace; it does not erase a logo, a new HTML table or a changed link. Conversely, a change consisting solely of tolerated whitespace need not invalidate that signature. Check any l= body-length limit when describing coverage, but do not propose leaving newly appended business content unsigned as the default repair. RFC 6376 sections 3.4 and 8.2

Compare MIME forms and route variants

A message can contain plain-text and HTML alternatives, embedded images and attachments. A server-side addition may alter more than the part displayed by one mail client. Capture unchanged raw messages at the relevant boundaries and compare MIME structure, encoding and the resulting canonicalized body. An image that looks identical on screen can still accompany a different serialized message.

Include the routes that matter to the organization: desktop composition, mobile sending, shared-address replies and automated applications where applicable. Their paths may differ. A pass on one desktop message cannot prove that a mobile message used the same signer or avoided a second insertion. Record unknown routes as gaps rather than inventing provider defaults to fill them.

Work through a fictional double insertion

Imagine fictional example.com prepares a message containing a user-added text signature. Its outbound signer applies a full-body DKIM signature with d=example.com. A later service appends a centrally managed HTML signature before external delivery. Assume the input signature verified and the added material changes the canonicalized body. The original signature then fails verification on the modified representation. This is a constructed example, not an executed trial.

Suppose the service also signs with d=signature-service.example while the visible From remains person@example.com. That new signature can be valid without aligning with the author domain. DMARC evaluates aligned authenticated identifiers, not whichever signature was added most recently. A separate aligned SPF result could still provide a passing path, so the failed original DKIM signature alone is not a complete DMARC diagnosis. RFC 9989 sections 3.2.10 and 5.3

Review signing order without granting new authority

For legitimate example.com outbound mail, a proposed integration might complete visible-content insertion before an authorized example.com signer processes the final message. Another supported architecture may have an appropriately authorized service perform aligned signing after its additions. Which option is available depends on documented product behavior and the actual tenant design; neither has been configured by this article. RFC 6376 section 5.1

Changing signing order is different from handing a vendor private-key material or broad relay permissions. Identify the responsible domain, approved identity and credential boundary before proposing a deployment. Adding a vendor to SPF does not repair a changed DKIM body hash, and it does not automatically make an unrelated envelope identity align. Keep envelope authorization and message-signature integrity as distinct tasks. RFC 7208 section 2.4 RFC 9989 section 3.2.4

Verify both appearance and authentication

The acceptance plan should include intended branding, absence of duplicate additions, correct links and readable plain-text output alongside cryptographic checks. A visually attractive result can fail authentication, and a valid signature can cover the wrong corporate footer. Assign separate owners for content approval and mail authentication so that neither criterion silently substitutes for the other.

After a separately authorized test, retain the original and final raw samples, route evidence, relevant public-key context and per-signature verifier results. Use Authentication-Results only from an identified trusted evaluator; a copied field is not proof of the final message's integrity. This package contains no such operational samples or completed account-specific validation. RFC 8601 section 7

[ Takeaway ]

Complete intended visible-content changes before the authorized final signer when the supported architecture permits it, and verify the actual path instead of assuming vendor behavior.

[ By role ]

What this means for you

Choose your role for practical next steps.

Branding owner

Define the intended final content

A cryptographic pass does not establish that the correct contact details and legal wording appeared.

Next steps

  1. Approve plain-text and HTML appearance separately.
  2. Check for repeated signatures across client and server insertion.
  3. Keep content acceptance separate from the authentication result.

Mail integration engineer

Locate insertion and final signing

The integration diagram must describe the deployed route, not just a generic vendor setup.

Next steps

  1. Identify the sender and client routes in scope.
  2. Verify any supported final aligned signer after legitimate changes.
  3. Keep domain authorization and credential custody explicit.

Email reliability analyst

Preserve per-signature evidence

Several signatures can yield different outcomes on the same modified message.

Next steps

  1. Record each signing domain and verifier result.
  2. Compare canonicalized MIME content at the relevant boundaries.
  3. Avoid declaring all clients ready from one successful desktop test.

[ FAQs ]

Frequently asked questions

Is the signature users see the same as the DKIM signature?

No. A visible signature is message content such as contact details or branding. DKIM is a cryptographic header covering selected message input. Their interaction depends on whether content insertion occurs before or after the evaluated DKIM signature.

Must a desktop and mobile message take the same signature-service path?

No such assumption is justified here. Client composition, server fallback and integration settings can differ. Establish the supported design and capture the actual route for each relevant sender and client before generalizing.

Does relaxed body canonicalization ignore a newly inserted logo?

No. It normalizes specific whitespace patterns rather than arbitrary HTML, images or new wording. Compare the canonicalized serialized body, including MIME structure, to determine whether this signature's covered input changed.

Can identical on-screen rendering conceal a DKIM-relevant change?

Yes. A rendered view does not fully expose MIME boundaries, encoding or raw representation. Preserve raw messages and evaluate the selected canonicalized input rather than deciding integrity from appearance alone.

Why can the service's DKIM pass still leave no aligned DKIM path?

Its d= domain may be unrelated to the author domain in From. DMARC requires the valid signing identity to align under the applicable mode. The newest signature is not automatically the relevant aligned signature.

Would adding the signature vendor to SPF fix a body-hash mismatch?

No. SPF authorizes a sending connection for an envelope identity; it does not recompute the DKIM body signature. A passing SPF result must also align with the author domain to provide the DMARC path under discussion.

Should the repair let newly appended content fall outside DKIM body coverage?

Not as a default workaround. Reduced body coverage can leave appended material unauthenticated. Prefer reviewing supported insertion and authorized signing order while retaining the intended content and security requirements.

What has to be proven before approving this integration?

Retain original and modified raw messages, actual signer placement and a before-and-after verification test, plus appearance and duplicate-insertion checks for relevant routes. None of those tenant-specific observations exists in this draft.

[ Sources ]

Sources

  1. Microsoft Exchange Online: Disclaimers, signatures, footers and headers
  2. RFC 6376: DomainKeys Identified Mail (DKIM) Signatures
  3. RFC 7208: Sender Policy Framework (SPF) for Authorizing Use of Domains in Email, Version 1
  4. RFC 8601: Message Header Field for Indicating Message Authentication Status
  5. RFC 9989: Domain-based Message Authentication, Reporting, and Conformance (DMARC)

[ Glossary ]

Glossary

Visible email signature
User-facing content such as contact details or branding inserted into a message, distinct from a DKIM-Signature header.
Server-side insertion
Addition of visible content by a service after submission rather than solely within the composing mail application; actual placement depends on the integration.
MIME representation
The serialized message structure and encoded parts that can differ from what a mail application visually displays.
Aligned signing domain
A DKIM d= domain from a validating signature that aligns with the author domain under the applicable DMARC mode.
Double insertion
A workflow in which more than one step adds visible signature content, requiring both appearance and authentication review.

Enterprise provenance

Why top financial institutions and governments retired their email security dashboards.

Traditional email authentication platforms sell you a dashboard and ask your engineers to do the labour. In high-stakes environments, manual oversight means delayed enforcement, configuration drift and unmonitored shadow domains.

EmailAuth is the agentic email provenance platform. Our agents enforce p=reject, keep third-party senders aligned and move your domains to full RFC 9989 compliance, the new DMARC standard,, with human-in-the-loop auditability.

Governments Large enterprises Tier-1 financial institutions

SaaS · Dedicated tenancy · Air-gapped on-prem · BYOK encryption

Knowledge Base