Brand impersonation: Internal Revenue Service

type.inbound
and (
  // display name contains IRS
  (
    strings.ilike(strings.replace_confusables(sender.display_name),
                  '*internal revenue service*'
    )
    or strings.like(strings.replace_confusables(sender.display_name), 'IRS*')
    or regex.icontains(strings.replace_confusables(sender.display_name),
                       'internal.{0,5}revenue.{0,5}service'
    )
  )

  // levenshtein distance similar to IRS
  or strings.ilevenshtein(strings.replace_confusables(sender.display_name),
                          'internal revenue service'
  ) <= 1
  or (
    strings.like(strings.replace_confusables(subject.base), '*IRS*')
    and any(ml.nlu_classifier(body.current_thread.text).topics,
            .name == "Government Services" and .confidence != "low"
    )
  )
  or 2 of (
    strings.icontains(body.current_thread.text, "Internal Revenue Service"),
    strings.icontains(body.current_thread.text, "4228 Park Ave S"),
    strings.icontains(body.current_thread.text, "1111 Constitution Ave"),
    strings.icontains(body.current_thread.text, "New York, New York 10003"),
    strings.icontains(body.current_thread.text, "Washington, DC 20224")
  )
  or regex.icontains(body.current_thread.text,
                     '©\s*20[0-9]{2}\s*\s*Internal Revenue Service'
  )
)
and (
  (
    any(ml.nlu_classifier(body.current_thread.text).topics,
        .name in ("Security and Authentication", "Financial Communications")
        and .confidence == "high"
    )
    and not any(ml.nlu_classifier(body.current_thread.text).topics,
                .name in (
                  "Advertising and Promotions",
                  "Newsletters and Digests",
                  "Political Mail",
                  "Events and Webinars"
                )
                and .confidence != "low"
    )
  )
  or (
    // OCR length is more than 2x the current_thread length
    // indicating that the body is mostly an image
    (
      (length(beta.ocr(file.message_screenshot()).text) + 0.0) / (
        length(body.current_thread.text) + 0.0
      )
    ) > 2
    and length(body.previous_threads) == 0
    and any(ml.nlu_classifier(beta.ocr(file.message_screenshot()).text).topics,
            .name in ("Security and Authentication", "Financial Communications")
            and .confidence == "high"
    )
    and not any(ml.nlu_classifier(beta.ocr(file.message_screenshot()).text).topics,
                .name in (
                  "Advertising and Promotions",
                  "Newsletters and Digests",
                  "Political Mail",
                  "Events and Webinars"
                )
                and .confidence != "low"
    )
  )
  or any(ml.nlu_classifier(body.current_thread.text).intents,
         .name == "cred_theft" and .confidence == "high"
  )
  or any(ml.nlu_classifier(beta.ocr(file.message_screenshot()).text).intents,
         .name == "cred_theft" and .confidence == "high"
  )
)
and not (
  (
    length(body.current_thread.text) > 2500
    or any(headers.hops,
           any(.fields,
               .name == 'List-Unsubscribe-Post'
               and .value == 'List-Unsubscribe=One-Click'
           )
    )
  )
  and any(ml.nlu_classifier(body.current_thread.text).intents,
          .name == "benign" and .confidence == "high"
  )
)

// and the sender is not in org_domains or from .gov domains and passes auth
and not (
  sender.email.domain.root_domain in $org_domains
  or (
    (
      sender.email.domain.root_domain in ("govdelivery.com", "ms-cpa.org")
      or sender.email.domain.tld == "gov"
    )
    and headers.auth_summary.dmarc.pass
  )
)
// and the sender is not from high trust sender root domains
and not (
  sender.email.domain.root_domain in $high_trust_sender_root_domains
  and coalesce(headers.auth_summary.dmarc.pass, false)
)