Build1 distinct publisher3 min readUpdated
Academics rebuilt Tiangou Secure Gateway firmware from leaked Git history and matched its filtering to the Great Firewall. One of three DNS injectors matched, and that ratio is the interesting part.
The Engineer · Build desk

Compiled by The EngineerSomething wrong?How this is made
A team of American academics told this year's USENIX security conference that the Chinese government is running Geedge Networks' Tiangou Secure Gateway as one of the Great Firewall's three known traffic filtering capabilities [1] [2]. They got there by reconstructing the device's firmware from more than 100,000 files that leaked out of Geedge's own network last year [3], which moves the story about Chinese censorship vendors from plausible inference to a named box with matched behaviour.
The leak was not a document dump. It included the databases behind Geedge's Jira and Confluence portals and all of the company's Git source code repositories, with commit history reaching back to November 2024 [4]. The researchers used that source and history to rebuild TSG firmware, then compared what the rebuilt filtering code does against observed sections and behaviour of the Great Firewall [5]. That is the part worth noting for anyone who does this kind of work: the comparison did not require physical access to a deployed appliance. A complete repository plus commit history is enough to produce an artifact you can characterise, and characterisation is what turns a censorship measurement into an attribution.
The scope claim deserves care. The newsletter reports TSG as one of three known filtering capabilities [2], while Bill Marczak noted on August 16, 2026 that Geedge is only one of the vendors involved and that the authors matched just one of the three characterised DNS injectors making up the Great Firewall to Geedge code [6] [7]. Two of the three injectors, on that account, still belong to somebody whose source code nobody has [8].
On engineering quality, the paper is blunt. The authors conclude TSG is as insecure and poorly coded as its Western counterparts [9], and write that its complexity and flexibility "are a double-edged sword": core components such as SAPP and its protocol plugins are written in memory-unsafe C, the system still leans on transitional arrangements like Stellar-on-SAPP, and code is copied from several third parties [10]. They say the Jira tickets corroborate that patchwork process, citing an AppSketch database upgrade that caused SAPP to restart as evidence of missing code verification [10]. The researchers expect the combination of published code, poor quality and abundant bugs to be usable by circumvention tools in future [11]. National filtering infrastructure procured as a commercial product inherits commercial defect rates, and a restart-on-upgrade bug in a censorship path is a gap in that path.
The same leak also exposed Geedge's export business, with reports shortly afterwards confirming sales of its censorship tooling to Kazakhstan, Ethiopia, Pakistan and Myanmar [12].
Watch whether the other two DNS injectors get attributed to other suppliers, because a three-vendor Great Firewall is a procurement structure rather than a single system, and each vendor is a separate leak risk [7] [8]. Watch whether circumvention projects publish anything that visibly exploits the reconstructed firmware [11]. And watch the export customers: if the same firmware family is deployed in Kazakhstan, Ethiopia, Pakistan or Myanmar, the reconstruction gives researchers a fingerprint to test for outside China [12].
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
A team of American academics found source code overlaps between the products of a Chinese tech company and China's Great Firewall traffic filtering and censorship system.
According to research presented at this year's USENIX security conference, the Chinese government is using the Geedge Networks Tiangou Secure Gateway (TSG) device as one of the Great Firewall's three known traffic filtering capabilities.
Researchers linked Geedge's device to the Great Firewall after more than 100,000 files leaked from Geedge's network last year.
The leak included databases for Geedge's Jira and Confluence portals as well as all its Git source code repositories, containing commit history going back to November 2024.
Researchers used the source code and commit history to reconstruct Geedge TSG firmware, which allowed them to match TSG traffic filtering capabilities to some sections of the Great Firewall and its behavior.
Bill Marczak (@billmarczak) noted on August 16, 2026 that Geedge is only one of the vendors involved in the Great Firewall.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Strong primary research, single secondary reporting channel
The underlying evidence is unusually concrete for an attribution story: a 100,000-file leak including full Git history, firmware reconstructed from that code, and behavioral matching against deployed GFW infrastructure, presented at USENIX and quoted verbatim on code-quality findings. It is capped because the cluster contains exactly one secondary source, the paper itself is not supplied, leak provenance is unverified in-cluster, and the named researcher caveat shows the match is partial (1 of 3 DNS injectors).
Confirmed state deployment plus named export markets, scale unquantified
Adoption of the technology itself is well attested: TSG is placed inside China's national filtering stack and reportedly exported to Kazakhstan, Ethiopia, Pakistan, and Myanmar. It is not higher because no capacity, contract, revenue, or per-country deployment figures appear, and the match to GFW components is explicitly partial.
Broadly aligned, mild over-attribution in framing
The reporting is close to its evidence and publishes the deflating caveat itself. The small positive gap comes from the headline framing of 'source code overlaps' and TSG as one of the GFW's three known filtering capabilities reading as fuller attribution than the 1-of-3 DNS injector match and the 'one vendor among several' qualification support, and from the circumvention-abuse angle being a researcher expectation with no demonstrated tooling.
Moderate: sponsored newsletter, academic publication incentives, no vendor rebuttal
Visible incentives are disclosed rather than hidden: the newsletter opens with a Socket Security sponsorship and promotes its own podcast, and the researchers have conference-publication and reputational incentives to present a clean attribution result. Offsetting this, the piece publishes a caveat that undercuts its own headline. The main unmeasured distortion risk is the absence of any Geedge or Chinese government response.
Moderate: solid method, single-publisher cluster
Confidence is limited chiefly by cluster composition rather than by weak underlying work: one publisher, one article, no supplied primary paper, and 'last year' / 'this year' time references that are not pinned to dates. The technical method, verbatim paper quote, dated researcher commentary, and self-published caveats support a moderate rather than low score.
security
Source-code match ties a shipping Chinese appliance to the Great Firewall1 distinct publisher
security
The Great Firewall has a vendor: Geedge Networks, five client states, 100,000 leaked files1 distinct publisher
product
Tor puts a censorship proxy on Google Play and makes bridge running a consumer install1 distinct publisher
security
The dependency gate moves upstream: why post-commit SCA misses hallucinated packages1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 21, 2026