Reconnaissance: All recipients cc/bcc'd or undisclosed

name: "Reconnaissance: All recipients cc/bcc'd or undisclosed"
description: |
  Recon messages, a form of deliverability testing, are used to validate whether a recipient address is valid or not, potentially preceding an attack.

  All recipients are bcc'd or undisclosed, with no links or attachments, and a short body and subject from an unknown sender.
type: "rule"
severity: "low"
source: |
  type.inbound
  and (
    length(recipients.bcc) > 0
    or length(recipients.cc) > 0
    or any(recipients.to, strings.ilike(.display_name, "undisclosed?recipients"))
  )
  and (
    length(subject.base) <= 10
    or (
      (
        strings.ilike(subject.base, "*checking*", "*testing*")
        or subject.base == body.current_thread.text
      )
      and length(subject.base) <= 25
    )
  )
  and length(attachments) == 0
  // and there are no links. Or all the links are to aka.ms or an extraction from a warning banner that match the senders domain
  and (
    length(body.links) == 0
    or length(filter(body.links,
                     (
                       .display_text is null
                       and .display_url.url == sender.email.domain.root_domain
                     )
                     or .href_url.domain.domain == "aka.ms"
              )
    ) == length(body.links)
  )
  and (
    body.current_thread.text is null
    or length(body.current_thread.text) < 50
    or (
      length(body.current_thread.text) < 900
      // or body is most likely all warning banner ending with a generic greeting
      and regex.imatch(body.current_thread.text, '.*(hi|hello)')
    )
    // body length without disclaimer is shorter than 50 characters
    or (
      any(map(filter(ml.nlu_classifier(body.current_thread.text).entities,
                     .name == "disclaimer"
              ),
              .text
          ),
          (length(body.current_thread.text) - length(.)) < 50
      )
    )
    // matching nlu_classifier 'bec' to smaller messages less than 200
    or (
      any(ml.nlu_classifier(body.current_thread.text).intents,
          .name == "bec" and .confidence in ("high", "medium")
      )
      and length(body.current_thread.text) < 200
      and not regex.icontains(body.html.raw,
                              '(?:<div data-smartmail=|gmail_signature(?:_[^"]*)?)'
      ) // not condition to exclude smaller messages with a legitimate email signature
      and length(body.previous_threads) == 0 // excluding messages with simple responses to other threads
    )
  )
  and profile.by_sender().prevalence != "common"
  and not profile.by_sender().solicited
  and not profile.by_sender().any_messages_benign
  
  // negate highly trusted sender domains unless they fail DMARC authentication
  and (
    (
      sender.email.domain.root_domain in $high_trust_sender_root_domains
      and not headers.auth_summary.dmarc.pass
    )
    or sender.email.domain.root_domain not in $high_trust_sender_root_domains
  )
tags:
  - "Attack surface reduction"
  - "Deliverability testing"
attack_types:
  - "Reconnaissance"
detection_methods:
  - "Content analysis"
  - "Header analysis"
  - "Sender analysis"
id: "420f60d3-5d10-5384-9253-9521a758e799"