sublimehighRule

Potential prompt injection attack in body HTML

Detects messages containing references to major AI tools (like Gemini, Copilot, ChatGPT, or Claude) in non-standard HTML elements.

MITRE ATT&CK

T1566.003 T1598 T1566 T1566.001 T1566.002 T1486 T1657 T1204.002 T1534 T1656 T1036 T1027

defense-evasioninitial-access

Detection Query

type.inbound
and length(filter(html.xpath(body.html, "//*[local-name() = 'admin']").nodes,
                  length(.display_text) > 0
                  and strings.ilike(.display_text,
                                    '*gemini*',
                                    '*copilot*',
                                    '*chatgpt*',
                                    '*claude*'
                  )
           )
) > 0

// 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
)

Data Sources

Email MessagesEmail HeadersEmail Attachments

Platforms

Raw Content

name: "Potential prompt injection attack in body HTML"
description: "Detects messages containing references to major AI tools (like Gemini, Copilot, ChatGPT, or Claude) in non-standard HTML elements."
type: "rule"
severity: "high"
source: |
  type.inbound
  and length(filter(html.xpath(body.html, "//*[local-name() = 'admin']").nodes,
                    length(.display_text) > 0
                    and strings.ilike(.display_text,
                                      '*gemini*',
                                      '*copilot*',
                                      '*chatgpt*',
                                      '*claude*'
                    )
             )
  ) > 0
  
  // 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
  )
attack_types:
  - "Callback Phishing"
  - "Credential Phishing"
  - "Extortion"
  - "Malware/Ransomware"
  - "Spam"
  - "BEC/Fraud"
tactics_and_techniques:
  - "Evasion"
  - "Social engineering"
detection_methods:
  - "Header analysis"
  - "HTML analysis"
  - "Content analysis"
id: "5fb24736-df8a-5a3a-84da-a2d5560d73d1"