Description

Detects messages impersonating the Canada Revenue Agency (CRA) in English or French that contain credential theft indicators. The rule identifies senders claiming to be CRA through display names or subject line references, uses natural language understanding to detect credential theft intent, and excludes legitimate senders with proper authentication.

References

No references.

Sublime Security
Created Jun 8th, 2026 • Last updated Aug 1st, 2026
Source
type.inbound
and (
  // cred theft
  any(ml.nlu_classifier(body.current_thread.text).intents,
      .name == "cred_theft" and .confidence != 'low'
  )
  // contains a link not to the domain "canda.ca"
  or any(body.current_thread.links,
         .href_url.domain.root_domain not in ('canada.ca')
  )
  // contains any attachment that isn't an image
  or any(attachments, .file_type not in $file_types_images)
)

// sender claims to be CRA
and (
  strings.icontains(sender.display_name,
                    'canada revenue agency',
                    'agence du revenu du canada'
  )
  or (
    regex.icontains(sender.display_name, '\bcra\b')
    // limit it to samples that mention CRA by name in the body
    and regex.icontains(body.current_thread.text,
                        '(?:canada revenue|revenu du canada)'
    )
  )
)

// not a high trust or cra-arc.gc.ca
and not (
  (
    // negate highly trusted sender domains
    sender.email.domain.root_domain in $high_trust_sender_root_domains
    // negate legit senders from legitimate cra
    or sender.email.domain.root_domain == "cra-arc.gc.ca"
  )
  // enforce authentication
  and coalesce(headers.auth_summary.dmarc.pass, false)
)
MQL Rule Console
DocsLearning Labs

Playground

Test against your own EMLs or sample data.

Share

Post about this on your socials.

Get Started. Today.

Managed or self-managed. No MX changes.

Deploy and integrate a free Sublime instance in minutes.
Get Started