• Sublime Core Feed
Medium Severity

Credential Phishing: Image as content, short or no body contents

Labels

Credential Phishing
Evasion
Image as content
Computer Vision
Content analysis
File analysis
Header analysis
Natural Language Understanding
Optical Character Recognition

Description

This rule identifies incoming messages with minimal links, all image attachments and either empty, brief or the body text is only a warning banner/disclaimer. It also checks for truncated PNG images or logos in addition to high-confidence credit theft intentions.

References

No references.

Sublime Security
Created Sep 8th, 2023 • Last updated Sep 8th, 2023
Feed Source
Sublime Core Feed
Source
GitHub
type.inbound
and length(body.links) < 2
and 0 < (length(attachments)) < 3
and (
  // body text is very short
  (
    0 <= (length(body.current_thread.text)) < 10 or body.current_thread.text is null
  )
  or (
    length(body.current_thread.text) < 900
    // or body is most likely all warning banner (text contains the sender and common warning banner language)
    and (
      (
        strings.contains(body.current_thread.text, sender.email.email)
        and strings.contains(body.current_thread.text, 'caution')
      )
      or regex.icontains(body.current_thread.text,
                         "intended recipient's use only|external email|sent from outside|you don't often"
      )
    )
  )
)
and (
  all(attachments,
      (.file_type in $file_types_images)
      and (
        any(file.explode(.),
            any(.scan.exiftool.fields, .value == "Truncated PNG image")
            or (
              any(ml.logo_detect(..).brands, .name is not null)
              and any(ml.nlu_classifier(.scan.ocr.raw).intents,
                      .name == "cred_theft" and .confidence == "high"
              )
            )
        )
      )
  )
)
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