Skip to content

Update: link_romance_suspicious.yml - #4964

Open
peterdj45 wants to merge 3 commits into
mainfrom
peter.fn.ESC-19445.link_romance_suspicious
Open

Update: link_romance_suspicious.yml#4964
peterdj45 wants to merge 3 commits into
mainfrom
peter.fn.ESC-19445.link_romance_suspicious

Conversation

@peterdj45

Copy link
Copy Markdown
Member

Description

adding logic to bypass NLU if the message is an automated form submission and has suspicious link terminology

adding telegraph file hosts (telegra.ph, graph.org) to suspicious link destination list

filtering condition that requires all links to not lead to the sender's domain to exclude .parser == plain

Associated samples

Enhanced detection rules for romance-related suspicious messages by adding checks for automated form submissions and explicit language in URLs.
@peterdj45
peterdj45 requested a review from a team July 27, 2026 23:41
@peterdj45
peterdj45 requested a review from a team as a code owner July 27, 2026 23:41
@github-actions github-actions Bot added the in-test-rules PR is in our testing suite to collect telemetry label Jul 27, 2026
github-actions Bot added a commit that referenced this pull request Jul 27, 2026
github-actions Bot added a commit that referenced this pull request Jul 28, 2026
Reduce FP surface on the form-submission NLU-bypass branch:
- Subject filter: strings.icontains -> regex.icontains with word-boundary
  \bform\b (was substring-matching information/platform/transform) and
  submi(?:t|ssion) stem
- URL-path regex: drop generic tokens (video, strip, cheating, affair,
  desire, intimate, explicit) that match benign CMS/webform link paths

Both associated samples still match; adult-specific tokens retained.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
github-actions Bot added a commit that referenced this pull request Jul 29, 2026
@peterdj45 peterdj45 changed the title Update romance suspicious detection rules Update: link_romance_suspicious.yml Jul 29, 2026
github-actions Bot added a commit that referenced this pull request Jul 29, 2026
The .parser != plain change surfaced FPs on Klaviyo-sent e-commerce
newsletters (Boredwalk, Disturbia, Fleur du Mal): their plain links point
to the sender domain while all hyperlinked CTAs route through Klaviyo
trackers (klclick.com/klclick1.com), so filtering plain links leaves only
off-domain trackers and the self-domain guard passes vacuously. The NLU
classifier separately misfires Romance/Sexually Explicit on marketing copy.

Negate when every clickable link is Klaviyo tracking AND the message
passes DMARC. Uses all() (not any()) so a single injected klclick link
can't bypass detection.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
github-actions Bot added a commit that referenced this pull request Jul 30, 2026
github-actions Bot added a commit that referenced this pull request Jul 30, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

in-test-rules PR is in our testing suite to collect telemetry

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant