Medium Severity

Attachment: Image-only docx/pptx callback phishing

Labels

Callback Phishing

Image as content

Impersonation: Brand

Social engineering

File analysis

Optical Character Recognition

Content analysis

Description

Detects inbound emails with docx or pptx attachments that contain no text but exactly one embedded image, a common tactic to evade text-based scanning by presenting content as a picture. The rule then inspects any extracted text from the image for a combination of callback phishing lures—such as references to subscriptions, invoices, refunds, or antivirus renewals alongside phone numbers or dollar amounts—paired with impersonation of well-known brands like PayPal, McAfee, Norton, or Best Buy, indicating a fraudulent phone-based scam attempt.

References

No references.

Sublime Security

Created Aug 21st, 2026 • Last updated Aug 21st, 2026

Feed Source

Sublime Core Feed

Source

GitHub

type.inbound
and sender.email.domain.root_domain in $free_email_providers
and any(attachments,
        .file_type in ("docx", "pptx")
        and any(file.explode(.),
                (.scan.docx.word_count == 0 and .scan.docx.image_count == 1)
                or (.scan.pptx.word_count == 0 and .scan.pptx.image_count == 1)
        )
        and any(file.explode(.),
                (.depth == 0 or .flavors.mime == "text/plain")
                // 4 of the following strings are found
                and 4 of (
                  // this section is synced with attachment_callback_phish_with_pdf.yml and body_callback_phishing_no_attachment.yml
                  strings.icontains(.scan.strings.raw, "purchase"),
                  strings.icontains(.scan.strings.raw, "payment"),
                  strings.icontains(.scan.strings.raw, "transaction"),
                  strings.icontains(.scan.strings.raw, "subscription"),
                  strings.icontains(.scan.strings.raw, "antivirus"),
                  strings.icontains(.scan.strings.raw, "order"),
                  strings.icontains(.scan.strings.raw, "support"),
                  strings.icontains(.scan.strings.raw, "help line"),
                  strings.icontains(.scan.strings.raw, "receipt"),
                  strings.icontains(.scan.strings.raw, "invoice"),
                  strings.icontains(.scan.strings.raw, "call"),
                  strings.icontains(.scan.strings.raw, "helpdesk"),
                  strings.icontains(.scan.strings.raw, "cancel"),
                  strings.icontains(.scan.strings.raw, "renew"),
                  strings.icontains(.scan.strings.raw, "refund"),
                  regex.icontains(.scan.strings.raw, "(?:reach|contact) us at"),
                  strings.icontains(.scan.strings.raw, "+1"),
                  strings.icontains(.scan.strings.raw, "amount"),
                  strings.icontains(.scan.strings.raw, "charged"),
                  strings.icontains(.scan.strings.raw, "crypto"),
                  strings.icontains(.scan.strings.raw, "wallet address"),
                  regex.icontains(.scan.strings.raw, '\$\d{3}\.\d{2}\b'),
                  regex.icontains(.scan.strings.raw,
                                  '\+?([ilo0-9]{1}.)?\(?[ilo0-9]{3}?\)?.[ilo0-9]{3}.?[ilo0-9]{4}',
                                  '\+?([ilo0-9]{1,2})?\s?\(?\d{3}\)?[\s\.\-⋅]{0,5}[ilo0-9]{3}[\s\.\-⋅]{0,5}[ilo0-9]{4}'
                  ),
                )
                and regex.icontains(.scan.strings.raw,
                                    '(p.{0,3}a.{0,3}y.{0,3}p.{0,3}a.{0,3}l|ma?c.?fee|n[o0]rt[o0]n|geek.{0,5}squad|ebay|symantec|best buy|lifel[o0]c|secure anywhere|starz|utilities premium|pc security|at&t|quickbooks|amazon)'
                )
        )
)

MQL Rule Console

•Docs•Learning Labs

Playground

Test against your own EMLs or sample data.

Post about this on your socials.

Get Started. Today.

Managed or self-managed. No MX changes.