The model directory
The Voidly Atlas ML stack
Censorship-intelligence models and evaluations in one place — forecast, classifier, anomaly, causal, and the accountability surfaces that grade them. Links to 34 records, and 8negative results we publish with the same permalinks as the wins.
Production defaults: classifier v3.3 + forecast v1 with per-country calibrators. Everything else is additive — new endpoints, not replacements. All CC BY 4.0.
Forecast models
| Model | What it does | Live |
|---|---|---|
| 7-day forecast | XGBoost + isotonic + per-country calibrators | API ↗ |
| Multi-horizon | 1d / 7d / 30d, per-horizon SHAP + conformal | API ↗ |
| Hourly | Within-day 6h / 12h / 24h horizons | API ↗ |
| 30-day trajectory | Seq2seq day-by-day path + bands | API ↗ |
| Per-region | Aggregate forecast per continent / MENA | API ↗ |
| Per-platform | Platform-level forecasts (Twitter, WhatsApp, …) | API ↗ |
| Per-domain | Domain-level forecasts for covered countries | API ↗ |
| Zero-shot transfer | Meta-feature model for low-data tail countries | API ↗ |
| Contagion chain | When X blocks, who follows in N days? | API ↗ |
| Live 24h watchlist | Proactive who-blocks-next ranking | API ↗ |
| Pre-shutdown signal | BGP / TLS / new-ASN precursors; see the dated evaluation | API ↗ |
Scroll to see every column.
Classifiers
| Model | What it does | Live |
|---|---|---|
| v3.3 (production) | GradientBoosting, regime-weighted contagion, classifier v3.3 — LOCO mean F1 0.7109 (0.63 on the 61 countries with n>=30); per-country performance varies | API ↗ |
| Per-country thresholds | F1-optimal thresholds per country; see the dated evaluation | API ↗ |
| Per-method | HTTP / TLS / DNS / TCP specialised | API ↗ |
| Per-protocol | Protocol-specific classifiers; inspect each validation split | API ↗ |
| Per-category | NEWS / ANON / COMT / … category classifiers | API ↗ |
| Per-measurement | Row-level classifier (Niaki KDD23) | API ↗ |
Scroll to see every column.
Anomaly detectors
| Model | What it does | Live |
|---|---|---|
| DBSCAN | CenDTect-style per-country shape anomaly | API ↗ |
| HDBSCAN domain-drift | Per-domain weekly blocking-pattern drift | API ↗ |
| STL seasonal | Deviation from a country’s own weekly pattern | API ↗ |
| Multi-country bursts | Coordinated-campaign detection, FDR-corrected | API ↗ |
| Fused ensemble | Detector fusion exposed for transparency; forward validation did not clear its promotion floor | API ↗ |
Scroll to see every column.
Causal & attribution
| Model | What it does | Live |
|---|---|---|
| Synthetic DiD | Shutdown attribution vs democracy donor pool | API ↗ |
| Causal forest HTE | Election treatment effect; in-sample association, not validated out of sample | API ↗ |
| Outage attribution | Censorship vs DDoS / infrastructure / weather | API ↗ |
| Survival / duration | Random Survival Forest; test discrimination is near random | API ↗ |
Scroll to see every column.
Accountability surfaces
| Model | What it does | Live |
|---|---|---|
| Prediction track record | Live 30-day out-of-sample accuracy per model | API ↗ |
| Baseline benchmark | Model lift over trivial baselines (honest) | API ↗ |
| Adversarial robustness | Detection rates under feature-level evasion tests | API ↗ |
| Serving reliability | Uptime + latency across the measured ML endpoints | API ↗ |
| Alert lead-time | Real Sentinel TP / FP rate + lead time | API ↗ |
| Calibration drift | Per-country forecast calibration audit | API ↗ |
| Model uncertainty | Per-day which predictions to question | API ↗ |
| Data freshness | Per-country A–D probe-freshness grade | API ↗ |
Scroll to see every column.
8 honest negative results
Models we built, evaluated, and did NOT promote. Each is a published finding with the full metrics. The headline lesson: stacking and transformer architectures did not beat the simpler regime-weighted GradientBoosting on cross-country generalisation.