The Surveillance Machine: A Catalogue of What Governments Actually Built

A darkened intelligence operations centre: a vast curved screen shows a world map laced with glowing data-flow arcs between continents, with fibre-optic cables converging across the floor and analysts silhouetted at their desks.

Article two in the BioJejak series on web security and encryption. The first article argued that most organisations quietly gave away control of their systems. This one shows the scale of what that control-less infrastructure is exposed to, and then catalogues the programs in detail.

Why this article

The first article in this series made a single argument: over twenty years we handed our data and our systems to third parties we do not control. The obvious next question is who sits on the other side of that arrangement, and how much they can actually see. This article answers it from the public record, not from speculation. Its purpose is first to give you a feel for the scope and breadth of state surveillance, so that the catalogue that follows reads as a picture of one coherent machine rather than a list of strange code names, and then to lay out the programs themselves. It is long and reference-like by design. Read the framing, then treat the catalogue as a map you can come back to.

Where the record comes from

In 2013 Edward Snowden, a contractor working inside the United States National Security Agency, copied a large body of classified material and handed it to journalists. Reporting followed in the Guardian, the Washington Post, Der Spiegel, the Intercept and others, from 2013 through 2018. The documents that were actually published, several hundred of them, are collected in the public Snowden Surveillance Archive, which in the edition used here holds 474 documents. They describe the signals-intelligence programs of the NSA and its "Five Eyes" partners: the United Kingdom's GCHQ, Canada's CSE, Australia's ASD and New Zealand's GCSB.

Two things are worth holding in mind throughout. The documents mostly date from roughly 2007 to 2013, so they are a snapshot that is now well over a decade old, and capability of this kind does not shrink with time. And they are only the published, redacted subset. The full picture was larger than what follows.

The scope, in plain terms

This is the part worth sitting with before the list. What the documents describe is not the targeted surveillance of a handful of named suspects. It is an architecture built to collect as much as possible, by default, and to decide what mattered afterwards. The phrase that captured the ambition, which appears in the Menwith Hill material in the archive, was "collect it all."

The breadth is easier to grasp by dimension than in prose:

The funding tells you it was not a fringe activity. The classified US intelligence "black budget" for fiscal year 2013, itself among the leaked documents, came to about 52.6 billion dollars, of which the NSA's share was roughly 10.8 billion dollars. This was a core, richly resourced function of the state.

The point for this series is simple. If you run on infrastructure you do not control, this is the environment that infrastructure sits in. It is documented, not imagined, and it is the backdrop against which every later article about owning your systems should be read.

How to read the catalogue

The catalogue is organised by function into sixteen sections and two appendices. Each entry gives the program name, the agency or directorate that ran it, a short description, detailed information drawn from the documents, and the originating published document, cited by its archive filename (date, outlet and a short slug) so you can locate it yourself.

A few cautions on precision. Many documents are partial or redacted, so a program's full scope is sometimes only implied. A single codeword can name a program, a database, a tool, a cover term for a particular cable, or a security compartment, and the text notes which sense applies. Where a capability is clearly described but carries no codename in the released material, it is listed in section sixteen under an assigned label (Unidentified Project N). The agencies, for reference: NSA (United States), GCHQ (United Kingdom), CSE or CSEC (Canada), ASD or DSD (Australia), GCSB (New Zealand), with "Five Eyes" meaning all five together. Use what follows as a map, not a cover-to-cover read.

1. Bulk collection and corporate access (Special Source Operations)

Special Source Operations (SSO) is the NSA division that runs collection through corporate partnerships and cable access. It is split into two lines: PRISM (data obtained from U.S. internet companies under legal process) and Upstream (collection from the internet backbone and cables).

2. Databases, analytics and metadata systems

3. Metadata sharing across agencies

4. Computer Network Exploitation: TAO, QUANTUM and the exploitation chain

Tailored Access Operations (TAO), later part of the Computer Network Operations (CNO) mission, is the NSA's hacking arm. Its Remote Operations Center (ROC) runs active operations. The broad initiative to put implants on large numbers of machines is GENIE; TURBINE automates the command and control of those implants at scale; TURMOIL is the passive sensor layer; together they sit in the TURBULENCE architecture.

5. The ANT catalog (TAO Advanced/Access Network Technology implants)

A single 2008-era catalog (published 30 December 2013) lists roughly fifty hardware and software implants and surveillance gadgets offered by the ANT unit to TAO operators. They are built to GENIE-compliant frameworks (CHIMNEYPOOL, FREEFLOW) and families such as STRAITBIZARRE and COTTONMOUTH. Source for all items in this section: 20131230-jacobappelbaum__nsa_ant_catalog (cross-referenced where noted).

6. Defeating encryption (BULLRUN and the CES exploitation chain)

7. Mobile networks, SMS and geolocation

8. FORNSAT and satellite collection (Menwith Hill and partner sites)

Foreign satellite collection (FORNSAT) intercepts the downlinks of commercial and other satellites. Menwith Hill Station (MHS) in England is the hub for much of what follows.

9. GCHQ (United Kingdom) programs

10. CSE / CSEC (Canada) programs

11. Network mapping and situational awareness

12. Other agencies and partner programs

13. Counter-intrusion and "fourth-party" collection

"Fourth-party collection" means the NSA watching, or riding on, another country's hacking so as to steal its take or use its tools.

14. Anonymity networks (Tor) targeting

15. Compartments, cover terms and classification systems

These are not collection programs in themselves; they are the control markings that protect the programs above. They appear throughout the archive and are worth cataloguing because they map the hidden structure.

16. Unidentified programs (described in the documents, no codename released)

Where the archive describes a distinct capability or operation but the released pages give it no cover term, it is listed here under an assigned label. The label is mine, for reference only; it is not an NSA codename.

Where this leads: taking control back

Read in one sitting, the catalogue can feel like a counsel of despair. It is the opposite. Almost every program in it depends on a specific, ordinary weakness: data handed to a third party, a cable left unencrypted, a standard quietly weakened, a device bought through a supply chain nobody checked, a phone that trusts its carrier. Each of those is a choice, and each can be made differently.

The third article in this series takes exactly that view. It sets out how an organisation, or a state, that takes its own security seriously would structure its communications to deny these programs their footholds: owning the hardware and the cryptographic keys, controlling the physical path, and facing the uncomfortable truth that the only channel none of this can reach is the one that is never electronic at all. If this article is the threat, the next one is the response.

Appendix A: Special Source Operations cover-term index

A quick cross-reference of the SSO "cover terms" (each masks a cable, access or partner), with their SIGADs where the documents give them.

Sources: 20131104-thewashingtonpost__sso_overview, 20131014-thewashingtonpost__sso_collection_optimization, 20180320-theintercept__second_look_sso_20mar2012_wstormbrewmap, 20161116-theintercept__sso_corporate_overview.

Appendix B: How this catalogue was produced

Text was extracted from all 474 PDFs with pdftotext; the 27 image-only documents were run through tesseract OCR. A concordance of capitalized codename candidates was built across the full corpus, ranked by how many documents each term appeared in, and then each candidate was read back in context in its source files. Codenames that are plainly OCR noise were discarded; terms that are classification markings or agency names (for example TOP SECRET, COMINT, NOFORN, SIGINT) were excluded from the program list and, where structural, summarized in section 15. Entries are grounded in the archive files cited on each line. Because some documents are redacted or are single pages pulled from larger originals, a few programs are described only as far as the released pages allow.

Previous article: How We Gave Away Control of Our IT. Part two of a three-part series.

Sources are the published documents of the Snowden Surveillance Archive (2013 to 2018), cited on each entry by date, outlet and slug. The documents mostly date from 2007 to 2013 and are a redacted, partial subset.

Comments

No comments yet.