Numbers Stations: What Is Actually Known

Tune across the shortwave bands for long enough and you will eventually find a voice reading numbers. Usually a woman, usually flat and unhurried, sometimes synthesised, reading digits in groups of five for twenty or forty minutes and then stopping. No station identification you can look up, no explanation, no acknowledgement that anyone is listening. They have been doing it for decades. A great deal is known about them and a great deal more is asserted, and the two sets overlap less than the internet suggests.

What a transmission actually sounds like

The format is consistent across stations and across decades, and it is not mysterious. Usually an interval signal — a few bars of music, a repeated tone, a spoken callword — running for some minutes so a listener can find the frequency. Then an identifier, often a three-figure group repeated. Then a count of how many groups are coming. Then the groups, read at dictation speed and very often read twice. Then an ending word and silence.

Every piece of that is ordinary radiotelegraph message procedure of the period. An interval signal to acquire the frequency, an address, a group count so the recipient can tell whether anything was lost in a fade, repetition for redundancy over a noisy channel, a defined end. It is the same discipline that governed maritime and military traffic. Nothing about the mechanics needs explaining. Only the context does.

The documented record

Hobbyist monitors have logged these transmissions systematically for decades — frequencies, schedules, formats, changes — and much of that record is public. The volunteer ENIGMA classification scheme gives each station a letter for the kind of transmission and a number: E for an English voice, G for German, S for Slavic, V for other languages, M for Morse, X for digital. A compilation of recordings released commercially in 1997 put a large archive in public hands. This is the least contested part of the subject: the broadcasts exist, they are on tape, and their behaviour over time is documented.

Lincolnshire Poacher

Classified E03. An English female voice reading five-figure groups, introduced by two bars of the folk tune it is named for. Logged from roughly the mid-1970s until 2008, when it stopped. Monitors placed the transmitter in Cyprus by direction finding; no government has confirmed it. A sister station using the tune Cherry Ripe ran Far East schedules and ended around the same time.

UVB-76, The Buzzer

A repetitive buzzing tone on 4625 kHz, documented since at least the early 1980s and still transmitting. Occasionally the buzz stops and a Russian voice reads a callword followed by names and groups. The identifier comes from an early announcement; others have been logged since. Direction finding places it in Russia. No official statement of purpose has ever been made.

HM01

A Cuban hybrid transmission monitored since around 2009, alternating spoken Spanish digit groups with segments of digital data. It succeeds a family of Spanish-language voice stations logged for decades.

The Morse ones

Voice gets the attention, but much of the same traffic has always gone out in Morse and, more recently, in digital modes. The format is identical. Only the modulation changed.

The one piece of hard evidence

Most attribution in this field is inference. There is one category that is not, and it is worth being precise about: numbers-station traffic has been entered into evidence in criminal proceedings.

In the United States federal prosecution of a Cuban intelligence network in Miami, decided in 2001, the government produced decrypted shortwave numbers messages in court, having recovered the means to decrypt them during searches. In a later Washington case, in 2009, a couple who pleaded to acting as agents of Cuba were described in the charging documents as having received instructions by coded shortwave transmission. Those are public court records, not anyone's theory, and they establish beyond argument that at least one family of these stations was an operational intelligence channel working exactly the way the obvious reading suggests.

What those cases do not establish is anything at all about any other station. Evidence about one transmitter is evidence about one transmitter.

Why a broadcast of digits is a sensible design

The cryptography is the easy part to explain, and explaining it removes most of the mystique.

A one-time pad is a key of truly random digits, as long as the message, used once and then destroyed. Claude Shannon proved in 1949 that such a cipher has perfect secrecy: the ciphertext reveals nothing about the plaintext except its length, no matter how much computing power an attacker has. It is the only cipher with that property and the constraints are the price of it.

In manual practice the plaintext is converted to digits with a small substitution table, then the pad digits are added one at a time modulo ten, without carrying. The recipient subtracts. The output is a string of digits — which is why the traffic is digits — broken into groups of five because that is what telegraph operators have always done to keep count. There is no cleverness in the format at all.

The channel choice follows from one property that is easy to overlook: the recipient never transmits. A listener with a radio emits nothing. There is no signal to direction-find, no callback, no registration, no account. Every other way of delivering a message — dead drops, couriers, telephones, the internet — creates something somewhere that can be watched. A shortwave receiver creates nothing.

The rest reinforces it. HF crosses continents without infrastructure inside the target country, so there is no domestic system to compromise and no metadata to subpoena. Broadcasting in the open to everybody conceals which of the millions of people in range the message is for, so traffic analysis yields nothing. A human voice reading digits slowly, twice, gets through fading that would defeat most data schemes of the era, and needs no equipment at the receiving end beyond ears, paper and a pencil. And because the cipher is a one-time pad, interception is worthless; the only useful thing to seize is the pad, which can be printed small enough to swallow.

That is also the answer to the question of why any of this survived the arrival of the internet. Nothing about the argument depended on the 1960s. It depended on the receiver being silent, and it still is. The physics that carries a shortwave signal across an ocean has not changed either.

Where the evidence stops

Well established. The transmissions exist and are recorded. The formats are consistent and documented by decades of independent monitoring. At least one station family was an intelligence channel, proved in court.

Reasonable inference. That these are state operations in general. The transmitter power, the schedule discipline maintained across decades, the frequency management and the sites themselves are beyond hobbyists or broadcasters, and no other kind of organisation has been proposed that could do it.

Not established, whatever you have read. Which agency operates any given station, in nearly every case. Enthusiast direction finding produces informed estimates of location, which is not confirmation of ownership and is not knowledge of purpose. The content of any message not decrypted with a recovered pad is unknown and, if the procedure was followed properly, unknowable.

Folklore. That the Buzzer is a trigger for an automatic retaliatory system has no evidence behind it and does not fit the station's observed behaviour. Reading world events into changes of schedule is pattern-seeking of the purest kind: a station transmitting continuously for forty years will always have done something unusual near any date you choose to test. Claims about submarines belong to a different and well-documented problem with a well-documented solution at far lower frequencies. And the idea that all this ended with the Cold War is simply wrong — several stations are transmitting today, on schedules anyone can check with a receiver.

A last piece of honesty in the other direction. Not every odd digit transmission on HF is espionage. Plenty of what gets posted as a numbers station is a channel marker, a frequency-management transmission, a military exercise or a misidentified utility service. The bands are full of things that are strange without being secret, which is a good description of shortwave in general and of the traffic that came off it.