This release deepens the analysis and makes the report easier to read. Each deduction is now explained by a concrete issue, attachments are scanned, BIMI is validated from the record down to the certificate, and most of the web interface has been reworked.
This release has been funded through NGI Zero Core, a fund established by NLnet with financial support from the European Commission's Next Generation Internet program. Learn more at the NLnet project page.
Highlights
- Every point is explained. Each check reports the points it deducts as an Issue, and the report shows how many points fixing each one would bring back.
- Attachment analysis. Attachments are inspected for malware, macros, active content and deceptive filenames.
- Full BIMI validation, including Verified Mark Certificates (VMC).
- EML upload. An existing
.emlfile can be analyzed without sending it to the test address. - A redesigned report. Checks are laid out as rows, sections that need no attention are folded, and the path of the email is drawn as a timeline.
- Domain reputation and sender-domain insights, along with a stricter parsing of SPF, DKIM and DMARC records.
New features
Attachment analysis
- Attachments are scanned with ClamAV, and their hashes are looked up on VirusTotal (the API key is shared with the other VirusTotal checks through a single flag).
- Executable content, Office macros, PDF active content, scripts and HTML smuggling are detected.
- Type mismatches and deceptive filenames are reported.
- Nested and password-protected archives are inspected, and attachments are extracted from embedded messages.
- Findings are also printed in the CLI text report.
BIMI
- BIMI validation now lives in a reusable
pkg/bimipackage, which can also be applied to an Assertion Record outside of any message. - Verified Mark Certificates are analyzed: the issuance chain is verified against the bundled Mark Verifying Authority roots, the certificate profile is checked, and the hash of the certified mark is verified.
- Logos are checked against the SVG Tiny Portable/Secure schema and for well-formedness. The file fetch is reported as evidence, SVGZ is supported, and an oversized logo is reported instead of being refused.
- Record discovery follows the organizational domain fallback and the local-part selector, and the
lps=andavp=tags are read. - An Indicator can only be displayed when DMARC is at enforcement, and the report now says so.
EML upload
- A new API endpoint and a web form accept an
.emlfile for analysis. Configuration flags control the feature, and they are documented along with the-emlCLI flag. - Reports now record their provenance, and the trusted
authserv-idis detected from theAuthentication-Resultsheaders. - The internal LMTP hops of the recipient are skipped when analyzing an uploaded message.
Sender and domain insights
- The origin of the sending IP (AS and country) is reported.
- The registration of the sender domain is looked up through RDAP: young and lapsing domains are flagged and charged.
- Disposable and free-provider sender domains are detected from embedded lists.
- Domain reputation is checked together with DNS, using the verdicts of checker-blacklist. It counts in the blacklist score and is shown on the domain page. Web ad blocklists are displayed but not scored.
- The blacklist feed caches can optionally be kept warm.
- The number of DNS lookups an SPF record costs is counted, and the record is displayed as a tree of its includes.
Content analysis
- The content checks are now a registry of criteria, and each finding carries its name, source, category and corroborations.
- The content of the message is submitted to rspamd for additional checks, and its symbols become findings.
- New checks:
- CSS features that some clients do not render (based on Can I email data),
- WCAG contrast of inline colors,
- a plain text part that differs from the HTML part,
- messages whose links all lead away from the sender,
- links and images served over
http:, - tracking pixels;
- Every URL of the message is fetched, and the result is reported.
- Link alignment compares texts and hrefs as A-labels, which makes it safe with internationalized domain names.
Spam filters and headers analysis
- Spam filters are graded on their verdict instead of on a linear scale. Verdicts stamped upstream are graded as well, and the Microsoft 365 verdict is shown among the filters (other third party verdicts will be supported in the future, please send us samples).
- Spam headers are attributed to the scanner that wrote them.
- The
x-return-mxverdicts of the receiver are displayed.
Improvements
Report and web interface
- Checks are laid out as rows of a
CheckList. ACheckCardsums up the issues of each check and folds the checks that pass. - Sections that need no attention are folded, whereas those with suggestions that cost points are opened. Each section shows its status under its grade.
- Authentication verdicts, header analysis, content analysis, domain reputation and spam filters are all presented as rows of checks. The authentication details show what led to each verdict, and domain alignment is detailed identity by identity.
- The email path is drawn as a timeline, starting with the sender.
- Findings link to the standards that define them.
- Folded content is opened when an anchor is followed, and history rows and pagination are real links.
- The interface adopts the HappyDomain colour theme and logo, a branded tile loader replaces the spinners, and the error page has been reworked.
- The whole frontend moved to SvelteKit 3.
- All the tools can be reached from the feature cards of the homepage.
- A warning is displayed when a report was generated long after the message was sent, and the age of the report appears in the summary.
- Categories that did not run (no sender IP, no spam filter, no blacklist check, no receiver checking the sender) are shown as N/A or left out of the overall grade, instead of displaying a misleading 0% or 100%.
Authentication and DNS
- Domain alignment follows the DMARC policy (relaxed or strict), reports the Return-Path address and is scored per identity.
- The includes of SPF records that use
redirectare followed, and such records are scored for what they are. - The SPF result checked against the HELO identity is kept, and the SPF policy of the HELO hostname is reported.
- The Authentication-Results field is parsed according to its grammar, in place of twenty patterns.
- DNS queries use absolute names, so that the search list of the resolver can never answer.
- A domain without MX is told apart from one whose MX lead nowhere, and a null MX is recognized.
- SPF softfail and DMARC
p=noneare reported as suggestions. - DNS checks are grouped by subject, without alerts.
- Messages are decoded with go-message (MIME transfer-encoding, charsets, uuencode).
Docker image
authentication_milternow ships a realmail-dmarc.ini, enables the HELO check and reports SPF upgrade attacks. Debug output can be enabled withMILTER_DEBUG.- The milter and spamd worker pools have been shrunk.
- Alpine is pinned to 3.24, and
spamass-milteris fetched from the Savannah mirror and tracked by Renovate.
Bug fixes
SPF
- Lookups are no longer counted past the
allmechanism, and terms that evaluation cannot reach are ignored. - SPF terms and the version term are read case-insensitively, the
allmechanism is found wherever it stands, and a record withoutallis read as ending in?all. - Repeated or empty
redirectandexpmodifiers are refused, and the addresses ofip4andip6mechanisms are checked. - Unknown modifiers are left to the receivers.
DMARC
- Tags are read regardless of case,
pis told apart fromspandnp, and a record that repeats a tag is refused. - No DMARC record is applied when a domain publishes several.
- An unknown policy is no longer taken for the strictest one, and reporting addresses that cannot work are reported.
DKIM
- A revoked key is reported, and a key ruled out by its
htag is refused.
BIMI
- A foreign TXT record is reported as no BIMI record.
- A declined BIMI is charged what it really costs, and all the points are granted only to a certified logo.
Scoring and analysis
- Links and HTML that are fine are no longer penalised, and links refused to automated clients are no longer counted as broken.
- Google's internal signature is charged only when it fails, and a single image without description costs at most three points.
- A defect seen by two checks is reported, and charged, only once.
- Network-wide blacklist listings are kept below the scoring ones, and a DNS list configured twice is queried once.
- The DNS A+ obtained from a HELO hostname without SPF policy is kept.
- PTR/FCrDNS is excluded from the DNS score when there is no sender IP.
- Comments in
MIME-Versionare ignored, and encoded redirect URLs no longer trigger the false positive on template placeholders. - The domain misalignment check on
httplinks is fixed.
Other
- The report page stays up when a spam filter gives no threshold.
- The size of the incoming message is also bounded on LMTP.
- Multi-line header matching in the receiver is fixed.
- The web UI shows report fetch errors instead of a blank page, retries on rate limiting, and no longer misses the app-config injection on routes served through the dev proxy.
- The summary wording for a missing SPF record is fixed, as well as the
Not testedandNot configuredlabels.
API changes
- A new endpoint accepts EML uploads, and the raw email route is now
/report/{id}/raw.eml. ContentIssueis renamedIssue. Issues now carry a category, a name and the points they cost.- New schemas describe the sender origin, the sender domain information and the domain reputation verdicts.