Datasets:

emperor-mew
/

ooni-censorship-historical

license: cc-by-4.0
task_categories:
  - tabular-classification
  - time-series-forecasting
language:
  - en
tags:
  - censorship
  - internet-freedom
  - ooni
  - human-rights
  - network-measurement
  - geopolitics
pretty_name: Voidly OONI Censorship Historical
size_categories:
  - 1M<n<10M

Voidly OONI Censorship Historical

A 10-year open archive for internet censorship research and ML.

Dataset Description

This dataset contains 10 years of global internet censorship measurements from 120+ countries:

1.6M+ daily measurements (2017-2026)
37K detected anomaly spikes
4.5K confirmed censorship events with labels
25+ known major incidents (Mahsa Amini protests, Myanmar coup, etc.)

Data Sources

Primary: OONI (Open Observatory of Network Interference)
Secondary: Voidly Research analysis and labeling

Files

File	Description	Rows
`data/ooni-historical.parquet`	Daily measurements by country/test	1.6M
`data/censorship-incidents.parquet`	Labeled anomaly spikes	37K

Usage

from datasets import load_dataset

# Load historical measurements
ds = load_dataset("emperor-mew/ooni-censorship-historical",
                  data_files="data/ooni-historical.parquet")

# Load labeled incidents (for ML training)
incidents = load_dataset("emperor-mew/ooni-censorship-historical",
                         data_files="data/censorship-incidents.parquet")

Schema

ooni-historical

Column	Type	Description
country	string	ISO 3166-1 alpha-2 country code
test_name	string	OONI test type (web_connectivity, telegram, whatsapp)
date	date	Measurement date
measurement_count	int	Total measurements
anomaly_count	int	Measurements showing anomalies
confirmed_count	int	Confirmed blocked
anomaly_rate	float	Fraction showing anomalies (0-1)

censorship-incidents

Column	Type	Description
country	string	ISO 3166-1 alpha-2 country code
date	date	Incident date
anomaly_rate	float	Measured anomaly rate
measurement_count	int	Sample size
spike_magnitude	float	Z-score above baseline
label	int	1=confirmed censorship, 0=not
event	string	Matched known event (if any)
confidence	float	Label confidence (0-1)

Known Events Covered

Iran Mahsa Amini protests (2022)
Myanmar military coup (2021)
Belarus election shutdown (2020)
Russia Ukraine invasion blocks (2022+)
Kazakhstan January protests (2022)
Sudan military coup (2021)
Cuba July protests (2021)
Uganda election shutdown (2021)
And 17+ more

Voidly Atlas ML Stack (2026-05-21)

This historical archive is the long-horizon training substrate for the Voidly Atlas ML stack. The production stack is documented in dedicated HuggingFace model cards under emperor-mew:

Classifier v3.3 (emperor-mew/voidly-classifier-v3.3) — country-day censorship classifier, GradientBoosting, regime-similarity-weighted contagion features. Honest cross-country generalization: leave-one-country-out median F1 0.87, mean F1 0.71. The fitted .pkl + per-country thresholds ship in that repo.
Multi-horizon forecast (emperor-mew/voidly-forecast-v1-multi-horizon) — 1d/7d/30d XGBoost + isotonic, LOCO AUC 0.91 / 0.88 / 0.84.
Unsupervised anomaly (emperor-mew/voidly-anomaly-dbscan-v1) — CenDTect-style DBSCAN second-opinion signal.
12 more model cards — search emperor-mew/voidly- on the Hub.

Note on the older "F1 99.8% / AUC 1.000" claim: that figure was a stratified-random-split number on a now-superseded v2 model. It does not reflect cross-country generalization. The current honest metric is the LOCO (leave-one-country-out) F1 reported above — random splits inflate apparent accuracy because the model learns per-country base rates.

For a clean held-out evaluation task, use the companion benchmark emperor-mew/voidly-bench-v1.

Citation

@dataset{voidly_ooni_historical_2026,
  author = {Voidly Research},
  title = {Voidly OONI Censorship Historical: 10 Years of Internet Measurement Data},
  year = {2026},
  publisher = {Hugging Face},
  url = {https://huggingface.co/datasets/emperor-mew/ooni-censorship-historical}
}

License

CC BY 4.0 - Attribution required