Skip to content

Network and Circumvention · Current evidence

Middlebox signatures register at 1.16% in the current OONI window

The decomposition separates interference signatures from transport failure and tests that could not execute.

1.16%middlebox index · across 6,147 completed middlebox tests

What the record says

Finding. The current measurement-weighted middlebox index is 1.16%.

This report is scoped to in-path-interference. It does not merge unlike instruments or infer a cause from co-movement.

How it was measured

rates are computed over COMPLETED tests only (measurement_count - failure_count). OONI's anomaly counter ranges only over completed runs, so a test that never completes has no anomaly rate — it has an execution-failure rate, reported separately

What this cannot establish

  • The index is a weighted aggregate over reporting tests; transport failure and execution blackout remain separate categories.

Instrument brief

live-reading · Current evidence

Palimpsest's view: in-path-interference is live. The current number is 1.2% over 6,147 completed middlebox tests. This is the instrument reading, not a national censorship rate.

Key number

  • 1.2% middlebox index · 6,147 completed middlebox tests

Current number and denominator

  • The current middlebox index is 1.2% over 6,147 completed middlebox tests.
  • The current measurement-weighted middlebox index is 1.16%.

Same-edition board context

  • Same-edition elevated layers: none declared. Other live lead or high signal ids: ddti, weibo-hotsearch, undertext, ooni-gfw, censored-planet, ioda-outages, vantage-fusion, erasure-observatory, wayback, app-storefront, apple-censorship, generative-firewall.
  • No editorial_priority or MAD score is treated as a finding on this reading.

What this reading does not show

  • The index is a weighted aggregate over reporting tests; transport failure and execution blackout remain separate categories.
  • The brief does not assign motive, identify a person, merge unlike denominators, or treat a missing reading as a zero.
  • rates are computed over COMPLETED tests only (measurement_count - failure_count). OONI's anomaly counter ranges only over completed runs, so a test that never completes has no anomaly rate - it has an execution-failure rate, reported separately

Elevated peers this edition

ddti (attention), weibo-hotsearch (attention), undertext (attention), ooni-gfw (network), censored-planet (network), ioda-outages (network), vantage-fusion (network), erasure-observatory (erasure), wayback (erasure), app-storefront (erasure), apple-censorship (erasure), generative-firewall (models)

Counterreadings

  • A current or withheld instrument value is not independent corroboration of any publisher article, and it is not a cause.
  • Co-movement with another live signal on this edition is layer context, not a shared rate or a coordinated action.

Limitations

  • The index is a weighted aggregate over reporting tests; transport failure and execution blackout remain separate categories.
  • The companion copies newsroom templates only; it adds no interview, article body, or generative prose.

Analysis instrumentv-322242d9a8b227702cb241fd · structured assessment · reading companion

Read the evidence

The exact source reading is in-path-interference-latest.json. The structured version of this dispatch is published beside the article.