Signal · Great Firewall, live
What the network firewall is blocking right now.
China's network censorship, read from outside the country. We ingest OONI's open measurements, taken by real probes inside China, and we run no probe of our own. A second instrument, Censored Planet, is shown alongside as an independent check. It is kept on its own denominator and never merged into a single figure.
Reading could not be loaded
This page publishes nothing when the feed does not load. There is no cached figure and no placeholder, because a plausible looking number here would be worse than a blank page.
Palimpsest is a free public good. MIT licensed, developed in the open, never a commercial product, and every finding on it ships the raw file it was computed from.
What funding buys Source on GitHubFinding 1 of 4
Pressure across the whole test panel
One measurement weighted figure over every OONI test method that produced Chinese data in the window. It answers a narrow question: across everything volunteers measured inside China, what share of measurements came back anomalous.
It is a pressure reading, not a count of blocked sites, and it moves with what volunteers happened to test as well as with what the censor did.
Great Firewall pressure index
The index across its whole recorded history
ooni-gfw-history.jsonlEvery committed reading of the index is published in the raw history file, and this chart draws from it when the page loads.
Nine weekly captures backfilled from OONI's archive open the record (May to early July, drawn as separate dots because they are separate captures), then live six-hourly runs from 11 July. The ◆ and ▲ marks are the committed timestamps of the only watch and only alarm this detector has ever raised, six days before China Digital Times published its documentation of that week's censored Zhengzhou-anniversary essays — the dated story is in DETECTIONS.md. A gap in the line is a gap in the record; nothing is interpolated.
loading
Per test evidence
loading reading
Finding 2 of 4
What was hit hardest
The domains with the highest anomaly rate in the window, each shown with the counts it was computed from. A domain measured twenty times and anomalous twenty times is not the same evidence as one measured twice.
This list reflects the URLs OONI volunteers actually tested. It is not a ranking of what China blocks most, and reading it that way would overstate it.
- loading reading
Finding 3 of 4
Two instruments, two denominators
OONI measures from probes run by volunteers inside China. Censored Planet measures remotely, using DNS and HTTP side channels from tens of thousands of vantage points outside it. Different technique, different population, different window.
Both detect Chinese interference. Their numbers are not comparable, and they are never averaged, added or presented here as one figure.
What the agreement buys is corroboration, not precision.
Instrument A · OONI
Instrument B · Censored Planet
Why these are never one number
loading denominators
Each figure is a share of a different thing. Combining them would produce a number that describes no measurement anyone made.
Finding 4 of 4
What people are reporting
The qualitative companion to the rates above: public issues opened on the net4people forum, where operators and researchers report blocking events and circumvention developments as they happen.
This is chatter, not measurement. It is useful for noticing a change before a rate moves, and it proves nothing on its own.
- loading
How each reading was made
Stated by the collectors themselves
- Loading
- loading
Why anomaly rate and not confirmed blocking
OONI marks a measurement confirmed when it sees an unambiguous block page. The Great Firewall rarely serves one. It injects a forged DNS answer or a TCP reset instead, so the connection simply fails and nothing announces itself as censorship.
Against China, confirmed sits near zero while real blocking is widespread. Tracking confirmed would report almost nothing. So this page tracks the anomaly rate, and states plainly that an anomaly is evidence of interference rather than proof of a policy decision.
We observe, we do not probe
Every figure on this page is computed from data somebody else already published. Palimpsest sends no request to a censored resource, runs no probe inside China, and asks no person there to do anything. These readings are vantage insensitive: running the collectors from anywhere produces the same numbers.
Watch the censor, never the censored.
Take the evidence with you
- ooni-gfw-latest.json The file the pressure index, the per test rows and the domain list are rendered from.
- ooni-gfw-history.jsonl One line per round, appended and never rewritten.
- censored-planet-latest.json The independent cross check, on its own denominator.
- net4people-latest.json The companion event log, ingested from public GitHub issues.
- collectors/ooni_gfw.py How the index is weighted, and when a test is reported as unavailable instead of given a rate.
- OONI open data The upstream source. Open data, reused with attribution.
- Observatory China censorship observatory Signals across the network, content and model layers, each publishing its own raw file. This page is one of them.
- Registry Verifiable eval registry Pre-registered suites, hash chained, anchored outside our own infrastructure. A peer of the observatory, not a side feature.
- Funding A funded public good MIT licensed and open from the first commit. Grants and donations pay for the collectors, the storage and the anchoring.
Data from OONI, open data, reused with attribution. Censored Planet data from data.censoredplanet.org. The companion event log is copyright net4people/bbs contributors, read from public GitHub issues. No individual measurement is attributed to anyone.