• Sublime Core Feed
High Severity

Credential phishing: Email delivery failure impersonation

Labels

Credential Phishing
Impersonation: Brand
Social engineering
Content analysis
Natural Language Understanding
Sender analysis

Description

Detects phishing emails impersonating email system notifications claiming delivery failures, rejected messages, or email system issues requiring user action to 'fix' or 'recover' email functionality. These attacks typically claim incoming emails couldn't be delivered and direct users to malicious portals to harvest credentials.

References

No references.

Sublime Security
Created Jun 6th, 2025 • Last updated Jul 16th, 2025
Feed Source
Sublime Core Feed
Source
GitHub
type.inbound
and length(body.links) < 10
and (
  any(ml.nlu_classifier(body.current_thread.text).intents,
      .name == "cred_theft" and .confidence == "high"
  )
  or (
    length(body.current_thread.text) < 250
    and any(recipients.to,
            strings.icontains(body.current_thread.text, .email.domain.sld)
            or strings.icontains(body.current_thread.text, .email.local_part)
    )
  )
)
and (
  regex.icontains(subject.subject, '(e)?mail(s)?')
  or (
    length(body.current_thread.text) < 700
    and strings.ilike(body.current_thread.text, '*mail*')
  )
)
and 3 of (
  strings.ilike(body.current_thread.text, "*incoming messages*"),
  strings.ilike(body.current_thread.text, "*server error*"),
  strings.ilike(body.current_thread.text, "*blocked*"),
  strings.ilike(body.current_thread.text, "*prevented*"),
  strings.ilike(body.current_thread.text, "*notification*"),
  strings.ilike(body.current_thread.text, "*fix email issues*"),
  strings.ilike(body.current_thread.text, "*rejected*"),
  strings.ilike(body.current_thread.text, "*recover and prevent*"),
  strings.ilike(body.current_thread.text, "*failure*"),
  strings.ilike(body.current_thread.text, "*rejection*"),
  strings.ilike(body.current_thread.text, "*failed*")
)
and (
  any(body.links,
      regex.icontains(.display_text,
                      "view",
                      "messages",
                      "recover",
                      "fix",
                      "portal",
                      "connect"
      )
      and not .display_text == "View Report"
      and .href_url.domain.root_domain in ("gmass.co")
  )
  or (
    length(body.links) < 3
    and any(body.links,
            any(recipients.to,
                .email.domain.root_domain == ..display_url.domain.root_domain
                and ..mismatched
            )
    )
  )
  or (all(recipients.to, .email.local_part == sender.display_name))
  or any(body.links,
         any(.href_url.rewrite.encoders, . == "proofpoint")
         and .href_url.domain.root_domain not in $tranco_50k
         and .href_url.domain.root_domain not in $org_domains
  )
)
and not any(body.links,
            regex.icontains(.display_text,
                            "view document",
                            "review (&|and) sign document"
            )
)
and sender.email.domain.root_domain not in (
  "bing.com",
  "microsoft.com",
  "microsoftonline.com",
  "microsoftsupport.com",
  "microsoft365.com",
  "office.com",
  "office365.com",
  "onedrive.com",
  "sharepointonline.com",
  "yammer.com",
  "ppops.net"
)

// negate org domains unless they fail DMARC authentication
and (
  (
    sender.email.domain.root_domain in $org_domains
    and (
      not headers.auth_summary.dmarc.pass
      // MS  emails from an org domain are router "internally" to MS, therefore, there is no authentication information
      or not (
        headers.auth_summary.dmarc.pass is null
        and all(headers.domains,
                .root_domain in ("outlook.com", "office365.com")
        )
        // typical emails from freemail Outlook accounts are from prod.outlook.com
        and strings.ends_with(headers.message_id, "protection.outlook.com>")
      )
    )
  )
  or sender.email.domain.root_domain not in $org_domains
)

// 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
)
and not profile.by_sender().solicited
and not profile.by_sender().any_messages_benign
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.

Get Started