Per-domain HDBSCAN drift leaderboard
Weekly unsupervised clustering of per-domain feature vectors (block rate, country footprint, dominant blocking method). Drift score = L2 distance between this week and last week in the same standardized space. A higher score means the domain's blocking footprint changed more.
Domains
—
Clusters this week
—
Clusters last week
—
New clusters this week
0
| # | Domain | Status | Last → This | Drift score | Block rate | n countries |
|---|---|---|---|---|---|---|
| 1 | stomp.straitstimes.com | new | C → C1 | — | 100.0% | 1 |
| 2 | www.tnp.sg | new | C → C1 | — | 100.0% | 1 |
| 3 | vpngate.net | new | C → noise | — | 0.0% | 0 |
| 4 | g-mens.net | new | C → noise | — | 0.0% | 0 |
| 5 | blog.fefe.de | new | C → noise | — | 90.9% | 1 |
| 6 | naenara.com.kp | new | C → noise | — | 10.0% | 1 |
| 7 | rargb.to | dropped | C13 → C | — | 100.0% | 10 |
| 8 | www.goldenpalace.com | dropped | C14 → C | — | 100.0% | 9 |
| 9 | scontent-frt3-2.cdninstagram.com | dropped | C3 → C | — | 100.0% | 10 |
| 10 | www.fuckingfreemovies.com | dropped | C13 → C | — | 100.0% | 10 |
| 11 | www.y2mate.com | dropped | C14 → C | — | 100.0% | 9 |
| 12 | xhamster.com | dropped | C14 → C | — | 100.0% | 9 |
| 13 | www.hrc.org | dropped | C14 → C | — | 100.0% | 9 |
| 14 | www.888casino.com | dropped | C14 → C | — | 100.0% | 8 |
| 15 | www.bravoporn.com | dropped | C14 → C | — | 100.0% | 8 |
| 16 | www.grindr.com | dropped | C14 → C | — | 100.0% | 9 |
| 17 | scontent.cdninstagram.com | dropped | C3 → C | — | 100.0% | 8 |
| 18 | www.4chan.org | dropped | C14 → C | — | 100.0% | 7 |
| 19 | www.gayscape.com | dropped | noise → C | — | 100.0% | 9 |
| 20 | beeg.com | dropped | C14 → C | — | 100.0% | 7 |
| 21 | www.onlinegambling.com | dropped | C14 → C | — | 100.0% | 7 |
| 22 | www.xnxx.com | dropped | C3 → C | — | 100.0% | 7 |
| 23 | www.youporn.com | dropped | C14 → C | — | 100.0% | 7 |
| 24 | pl.wikipedia.org | dropped | C1 → C | — | 0.0% | 0 |
| 25 | discord.com | dropped | noise → C | — | 100.0% | 7 |
| 26 | scontent-vie1-1.cdninstagram.com | dropped | C3 → C | — | 100.0% | 8 |
| 27 | staticxx.facebook.com | dropped | C3 → C | — | 100.0% | 7 |
| 28 | www.bglad.com | dropped | C14 → C | — | 100.0% | 8 |
| 29 | 1337x.to | dropped | C14 → C | — | 100.0% | 7 |
| 30 | bi.org | dropped | C14 → C | — | 100.0% | 7 |
GET /v1/anomaly/domain-drift/leaderboard
Raw JSON of the same data
HDBSCAN drift finding
Methodology, weekly cron schedule, honest caveats about week-over-week stability
Why a separate domain-drift surface? The per-country DBSCAN at /atlas/anomaly catches country-day shape anomalies. HDBSCAN at the domain level catches global blocking-pattern changes — “is anyone blocking openai.com differently this week?” The two are orthogonal: one detects per-country novelty, the other detects per-domain novelty.
Generated: 2026-08-16T04:00:06.178785Z