A censor can defeat the anomaly detector without triggering an alert by replacing blocked user traffic with synthetic requests from adversary-controlled machines, keeping per-jurisdiction connection counts within the typical range. The paper explicitly identifies this as an unaddressed active-attack vector.

Graduated censorship — limiting the suppression rate to remain within the typical weekly variance band — evades the weekly-interval detector entirely. The paper acknowledges that detecting slow-ramp blocking requires extending the observation window beyond seven days.

§6 detection

Per-jurisdiction user counts are modeled as a Poisson process; the detector infers the 99.99th-percentile credible interval for the underlying rate λ from the observed count via a Gamma-Poisson approximation rather than a Gaussian assumption, correctly treating small-jurisdiction zero-user days as non-anomalous.

§3–§4 defense

The detector constructs its 'typical ratio' baseline exclusively from the 50 largest jurisdictions, then discards outliers beyond four inter-quartile ranges of the median before fitting N(m,v). This ensures a jurisdiction undergoing active censorship cannot bias the global model and mask its own anomaly.

§4 defense

The deployed system uses 7-day intervals and a baseline built from the 50 largest Tor jurisdictions; a jurisdiction's user-count ratio is flagged when it falls outside the 99.99th percentile of the fitted Normal distribution N(m,v), yielding an expected false-alarm rate of approximately 1 in 10,000 per jurisdiction-week.

§5 evaluation

The largest single source of censored domains in the GNL is MESA lab's SNI monitoring dataset (E21-SNI-Top200w.txt) containing 57,362 censored domains, and E21-SNI-Top120W-20221020.txt with 36,467 domains—totaling over 93K domains from network tap data alone for a single country (E21 = Ethiopia per InterSecLab attribution). A separate Xinjiang dataset (XJ-CUCC-SNI-Top200w.txt) contains 13,604 domains. These datasets "do not seem to come from popular domain lists, and instead appear to be gathered from network taps," confirming that Geedge builds censorship target lists directly from passive traffic observation.

2026-sheffey-geedge §4.2, Table 3 dpisni-blockingmeasurement-platform

Of 6,915,266 domains extracted from the 572 GiB Geedge Networks Leak (GNL), 298,955 censored domains (93.7% of all GNL-censored domains) appear in neither Tranco top-1M nor CitizenLab test lists. Measurements across China (Guangzhou/Nanjing), Myanmar, Pakistan, and Algeria confirmed censorship via DNS injection and SNI-based TLS connection termination. The GNL covers 25–62% of Tranco-censored domains across countries, showing substantial but incomplete overlap. This vendor-side ground truth reveals a censorship surface roughly two orders of magnitude larger than curated academic test lists.

2026-sheffey-geedge §4.1, Table 2 dpisni-blockingdns-poisoningmeasurement-platform

CensorAlert aggregates censorship signals from heterogeneous open sources — including OONI, Cloudflare Radar, NetBlocks, GitHub Issues of circumvention tools (Hysteria, Xray), Net4People BBS, NTC Party forum, Mastodon, X, Telegram channels, and arXiv — normalizing each item into a common schema preserving timestamp, source, raw data, and provenance with a link to the original content.

2026-zohaib-extended §2 (Data Sources) measurement-platform

Systematic measurement platforms (OONI, Censored Planet, Cloudflare Radar, NetBlocks) have inherent blind spots due to geographic coverage and protocol-specific test constraints; critical censorship discoveries — including GFW fully-encrypted protocol blocking and regional Chinese censorship — were first surfaced by user reports on forums and GitHub issue pages of circumvention tools, not by automated measurement infrastructure.

2026-zohaib-extended §1 measurement-platformfully-encrypted-detect

CensorAlert generates text embeddings (OpenAI text-embedding-3-small) from each item's summary, title, and tags to cluster semantically similar reports within configurable time windows; near-duplicates such as reposts, copied headlines, and translations are collapsed into a single canonical post that preserves all original source URLs and metadata.

2026-zohaib-extended §2 (AI-based Scoring) measurement-platform

CensorAlert's LLM agent scores each ingested item 0–5 on five independent dimensions — credibility, novelty, impact, timeliness, and verifiability — then computes a normalized significance score (0–10) from these components; items are processed every two hours using OpenAI GPT-5 Thinking (hosted on Azure) constrained to return structured JSON output.

2026-zohaib-extended §2 (AI-based Scoring) measurement-platform