demo · v146

Intervention policy graph

Selective interventions don't fire on every permission prompt — Chrome scores the site against signals (abusive-notification report rate, deceptive-content flag, recency of user grant) and intervenes per API. Tune the signals and watch the graph rebuild the verdict per API.

The model used here mirrors the Safe Browsing & Permissions team's blog-post heuristic: site reputation × API-specific cost × user history. The bar shows the "risk score" per API; the verdict column shows whether Chrome grants normally, prompts more aggressively, or auto-blocks.

Site signals

User history

// Heuristic from the Permissions team blog post:
score = base[api]
      + 1.6 * abuseReports
      + 14  * deceptiveFlagged
      + 9   * shadyTld
      - 4   * recentGrants
      + 3   * dismissals
      - 0.4 * engagementVisits;
verdict = score > 70 ? "blocked"
        : score > 35 ? "prompt"
        : "granted";

see also