Bombes, Cribs and Colossi
- Codebreakers: The Inside Story of Bletchley Park edited by F.H. Hinsley and Alan Stripp
Oxford, 321 pp, £17.95, August 1993, ISBN 0 19 820327 6
Just before the outbreak of war the Government Code and Cypher School (GC & CS) moved from London to Bletchley Park in Buckinghamshire. In 1939, some hundred or so people were working there; by the end of the war there were about seven thousand. This book contains reminiscences by 30 of them, describing what they did and what life was like at Bletchley.
GC & CS came under ‘C’, the head of the Secret Service, and was nominally a department of the Foreign Office. In the years between the wars it had grown out of a group of code-breakers who, in World War One, had worked in the famous Room 40 at the Admiralty. Some of that group were still active in 1939; in particular Oliver Strachey, ‘Dilly’ Knox (whose work on the Abwehr version of Enigma is described in the book) and Commander Denniston, who became the first head of Bletchley. Already in 1938, Denniston had started to seek out the mathematicians, scholars and chess champions who reported to Bletchley on the outbreak of war. This slightly haphazard method of recruitment continued throughout the war. A number of the authors tell how their Oxford or Cambridge tutor (or friend of their tutor) asked them if they would like to do secret work, and how, at an interview in London, they were asked if they played chess, did crosswords or knew foreign languages. Those reading Classics were asked if they would like to learn Japanese. There were also experts in modern languages (two were known as ‘the walking dictionaries’) and, by 1943, many mathematicians (six of whom were, or were to become, fellows of the Royal Society). A Lancashire businessman, Wing-Commander Eric James, praised here for his skill in dealing with ‘tiresome intrigues’ and ‘something like chaos’, eventually got things running smoothly; after the war he became head of GCHQ at Cheltenham.
Enciphered enemy messages, broadcast in morse or teleprinter code, were sent to Bletchley by teleprinter or despatch rider from a chain of intercept (or ‘Y’) stations. Co-operation between these and Bletchley was essential, though not always straightforward. At the beginning the rule was that all decoding should be done at Bletchley; but much of the traffic was in low-grade hand cyphers, and it was gradually realised that this could best be decoded by Bletchley-trained people at the ‘Y’ station. This was particularly true of Japanese traffic. The important ‘Y’ stations for this were in East Africa, India and the Far East, and one of the book’s contributors, Hugh Denham, says that the work done at Bletchley on Japanese naval messages was almost without value.
At the RAF intercept station at Cheadle, there was a group of four ‘computors’, so called because officially their job was to decode air-to-ground traffic using daily keys provided by Bletchley: in fact, they often found the keys before Bletchley did. The computors believed the Air Ministry’s estimate of the size of the German bomber force in the west was much too high – Peter Gray Lucas refers to this as the ‘St Ives story’ (‘As I was going to St Ives, I met a man with seven wives’). But to get a better estimate it was necessary to know which aircraft belonged to which unit, which was difficult because the call-signs (three-place alpha-numeric) were continually changing; analysis at Bletchley had not found any discernible pattern. One night when traffic was light, two of the Cheadle computors sorted out the messages not by statistical means, but according to the style apparent in the choice of call-sign and found that this gave a coherent classification. Subsequently they were able to show that the size of a unit was nine aircraft, not 12 as the Air Ministry had supposed. Lucas gives this as an example of the ‘creative anarchy’ which contributed so much to Bletchley’s success.
The messages most copiously available were those encoded by the Enigma machine; in the last year of the war there were some 84,000 of these a month. About one third of Codebreakers is about the decoding, translation and evaluation of Enigma messages. Enigma was used by a large number of different German networks: maybe as many as a hundred, each with its own rules of procedure and daily keys. By 1943, decrypts were being mass-produced. Many messages were routinely broken, perhaps a few hours after the keys were changed, at midnight; others were never broken.
Enigma was a modification of a commercial machine; it was first adopted by the German Navy in 1926 and taken up by the Army and Air Force a few years later. As time went on, further modifications and complications were introduced. In the original form the keys were connected in typewriter keyboard order – Q W E R T Z U – to studs on a ring, current from which then passed through three rotors to a ‘reflector’, and back through the rotors to a lampboard, each lamp on which showed a particular letter. All the rotors, and the reflector, were internally wired so as to swap letters pairwise. This meant that no letter could code itself – probably a selling point for the original machine, but one of help to the codebreakers. Each rotor carried a ring marked with the letters of the alphabet, which could be rotated. The three current letters appeared at three windows, and each time a key was depressed, one or more of the rotors moved on. Already before the war a plugboard had been added, so that the sockets for, say, A and J could be connected by a cable, swapping the letters A and J both between the keyboard and the rotors, and between these and the lamp-board. By late 1940, the number of cross-pluggings was standardly ten. This meant that there were over 1014 (a hundred thousand billion) ways in which the plugboard could be set up.
A daily setting for the machine specified which rotors (out of a possible five) should be placed in the three positions, how the letter rings for these rotors should be set and which connections should be made on the plugboard. Monthly lists for the daily settings (supposedly chosen at random) were sent out to all users of a given network. When a message was to be sent, the operator chose, also supposedly at random, three letters for the text setting. After a preamble in clear he then used the machine, in some pre-arranged way, to encipher his text setting; then he set his rotors to it and enciphered the message. I have given a typical procedure; there were many variants.
Peter Twinn describes how, in February 1939, he assisted Dilly Knox in trying to break the Enigma code. They had some knowledge of the machine, having obtained the clear and cypher text, and plugboard setting, for a message sent in 1938. They knew that the order of the letters on the ring was not Q W E R T Z U; had they known what the order was they could have used their material to good effect. But in fact they were ‘like a couple of Micawbers’ waiting for something to turn up. What turned up was a meeting in July 1939 with some Polish cryptanalysts, who knew the right order of letters. Back in the early Thirties the Poles had guessed that, Germans being very orderly, it might be A B C D: it was. They also knew how the rotors were wired and had constructed facsimile machines, one of which they sent to GC & CS in August 1939. The Poles’ methods were based on permutation theory and some of them were to be rediscovered at Bletchley; others were passed on at the July meeting or later, when the Polish group had been installed in France. (These methods depended on a procedure for enciphering the text setting which the Germans abandoned in April 1940.) Stuart Milner-Barry comments on the sad and mysterious fact that after the fall of France the Polish group was not brought to Bletchley.
I have used the phrase ‘supposedly random’ because Jack Good emphasises here the difference between ‘random’ and ‘humanly random’. For example, operators might choose three successive letters of the alphabet on the keyboard, and sometimes the monthly list of settings was identical with one that had been used before. At Bletchley there were those who were quick to spot, or to guess at, such sloppiness, and to cash in on it.
There were other ways of arriving at possible conclusions about the settings. Good describes one, called ‘Banburismus’, in some detail. It involved sliding sheets of paper (made in Banbury), with perforations representing encrypts, one over the other: ‘The game of Banburismus was enjoyable, not easy enough to be trivial but not difficult enough to cause a nervous breakdown.’ For this and other methods, Alan Turing developed a highly original form of statistical theory which made it easy to calculate the relative weight of evidence supporting various hypotheses.
The most fruitful tool for code-breaking was a ‘crib’: a piece of clear text believed to correspond exactly to part of a ciphered one. A rather cumbrous bit of German wording (‘An Obergruppenführer’), standard phrases, captured material and the fact that messages (weather reports in particular) were sent out both by Enigma and in clear, or in some low-grade cypher, often allowed cribs to be found or guessed at. However, if you tried to use a crib in a simple-minded way you would have to go through an astronomical number of possible rotor and plugboard settings. Brilliant insights on the part of Gordon Welchman and Turing made it possible to design large ‘bombes’ – calculating devices which made such a search practicable. Given a ciphered text and a crib of, say, twenty to thirty letters, the first task was to select certain letters in the message to form a ‘menu’. The bombe had 12 simulated rotor assemblies, coupled together in the appropriate relative positions; thereafter they moved synchronously, step by step. The inputs and outputs of the rotor assemblies were connected, according to the menu, to a ‘diagonal’ board, allowing all the logical consequences of any assumption about the plugboard setting to be revealed electrically and instantaneously. Turing’s contribution was to show how an inconsistent assumption could be immediately rejected by a process analogous to the logical principle ex falso quodlibet – anything can follow from a contradiction. If the position of the rotors did not correspond to the text setting this would rapidly be recognised and the rotors would automatically be moved on; they stopped (and the operator was notified) if the setting was correct, or at least highly consistent with the crib. If the bombe had been set up assuming a wrong rotor order, this would show after it had run for about half an hour; it would have to be set up again by hand.
The bombe was important not only because messages could now be rapidly decrypted, but also because the ideas behind it seem never to have occurred to the German cryptologists. From time to time, despite the many precautions the Allies took to protect the source of their intelligence, the Germans wondered if Enigma was being broken and questioned their cryptologists. But the answer was always: ‘It is unbreakable.’ This confidence rested, presumably, on the fact that there were over 1014 plugboard settings.
The first few bombes started work in mid-1941; by the end of the war there were more than a hundred, each one the size of a very large bookcase. Diana Payne recalls her life tending one. She describes the finicky process of setting up a bombe (particularly when changing rotor positions) and her satisfaction at doing an important, though mysterious, job well.
In 1941, the Germans introduced a new kind of mechanical encoder called the Geheimschreiber; there were several variants, known collectively as ‘Fish’, but Bletchley concentrated on the one used by the Army, nicknamed ‘Tunny’. The book contains three descriptions of this machine, with some hints on how its code was broker – much of which information is new. ‘Tunny’ had a teleprinter keyboard; when a key was struck, a sequence of five bits (1 corresponding to ‘mark’ and 0 to ‘space’) was produced, according to the international teleprinter code. Simultaneously, a set of 12 wheels produced another group of five bits, which was added to the original sequence. The result was the encoded character, which could be sent out on line to a radio transmitter. The way in which the wheels (which might rotate at each step) produced the key was complicated. As with Enigma, there was a daily setting, and for each message a text setting. But by a lucky accident, two long messages were sent out with the same setting. Colonel Tiltman, the chief cryptanalyst and a brilliant one, disentangled the two messages and the key, and from this, after much hard work, a bright young mathematician calculated how the wheels worked. No one at Bletchley actually examined the German machine until after the war.
When settings were known a simulated machine was needed, to do the decoding; it was constructed using standard Post Office equipment. Some settings were broken by hand, but the mathematician M.H.A. Newman realised that what was needed was high-speed binary calculators. His section, the ‘Newmanry’, planned a series of such machines which were designed and constructed by Post Office engineers. By the end of the war ten ‘Colossus Mark II’ machines were at work: the very first functioning, large-scale (1500 valves) electronic computers. The importance of ‘Fish’ was that its traffic was strategic, (mostly from Berlin to army commands), rather than tactical, and still valuable even if the decoding took several days.
The breaking of Enigma and Fish was the most glamorous activity at Bletchley; but much other work went on there, and at intercept stations, breaking lower-grade cyphers by hand. The most commonly used of these was the enciphered code book. A code book assigns, arbitrarily, groups of four or five letters (or numbers) to commonly used words and phrases; the coded message is then enciphered using a random key taken from a page of the cypher book, the page number being somehow concealed in the transmitted encrypt. All the Japanese codes described here were of this kind. Breaking such a code involves, first of all, recovering the coded message and then reconstructing the code book. This is possible provided there is plenty of traffic and the books remain unchanged for long periods. Bit by bit good guesswork allows both operations to be done.
In peacetime, the function of GC & CS had been simply to decode; the assessment, even perhaps the translation, of the decrypts was to be done by the service intelligence departments in Whitehall. With the move to Bletchley it became obvious that some things were best done on the spot, and intelligence sections were set up. As time went on, these sections became more important, communicating directly with army commanders in the field, ensuring inter-service co-operation and deciding priorities for code-breaking. By the end of the war the ‘index’ at Bletchley was the ultimate authority on the organisation of the German services and their orders of battle. Churchill got better information about the progress of the war in Russia from Bletchley than he did from the Russians.
To begin with, however, the service intelligence departments, especially the Admiralty, treated Bletchley (and perhaps each other) with suspicion and contempt. The late Christopher Morris worked on low-grade naval cyphers and in April 1940 deciphered a message to merchant ships bound for Bergen, requiring them to report their positions to the German War Office: he was told that ships do not report to war offices and the message was disregarded. In fact, the ships were carrying troops for the invasion of Norway. F.H. Hinsley, who worked on naval traffic analysis and was alert to the slightest clue, describes how at first the Admiralty would not take seriously evaluations made by someone who had never been at sea or worked in Whitehall. Later they took him to sea (and to the Admiralty), and collaboration became full and effective.
In his Introduction Hinsley discusses the impact of code-breaking on the war in the west; in particular its use (and occasional non-use) in the Battle of the Atlantic, the war in North Africa and the preparations for D-Day. He draws attention to the fact that there were periods, sometimes of many months, when a particular sort of Enigma or Fish traffic could not be broken. Though cautious in his conclusions, he suggests that if Enigma and Fish had not been broken D-Day might have had to be delayed until 1946.
Most of the contributors have happy memories of their time at Bletchley – except one. Carmen Blacker was a young and promising student of Japanese; she was recruited by Bletchley and set to compiling, from miscellaneous captured material, an index of technical terms. She was isolated, poorly paid and miserable. She doubts that her index was ever useful. When, early in 1945, she wangled a transfer back to the School of Oriental Studies, her boss at Bletchley, an empire-builder, forbade anyone in the section to talk to her. Others recall the ‘warm and friendly atmosphere’, the ‘camaraderie’, the excitement of the daily breaks. It was out of the question to talk of one’s work outside Bletchley, or even outside of one’s own group. Some found this silence oppressive, but it also served as a bond. The common purpose and the need to work with each other, made the sections – particularly those breaking codes – democratic. Uniforms were seldom worn, rank was unimportant – a fact which sometimes upset high-ranking visitors. According to Morris, someone who wrote an intelligence report criticising Admiral Doenitz was reprimanded by the Admiralty and told that under no circumstances should a junior officer criticise an admiral. But it was precisely the lack of hierarchy and the encouragement of independent thought – the creative anarchy – that made the code-breakers so successful.