Gordon Welchman Read online

  During the autumn of 1941 the Whitehall intelligence departments had become increasingly concerned about BP’s control over intelligence. In particular, they were annoyed by BP’s right to communicate directly with fighting commands, even though it speeded up the process of getting vital information to commanders in the field. Inter-service rivalries had also surfaced in Hut 3 between the senior liaison officer with the Air Ministry, Robert Humphreys, backed by C. R. Curtis, head of the Army section and the nominal head of Hut 3, Malcolm Saunders. Humpheys had lobbied effectively for BP in Whitehall but was a poor team player and had caused great dissent in both Hut 3 and Whitehall. Nigel de Grey described the situation as ‘an imbroglio of conflicting jealousies, intrigue and differing opinions’. Eventually, an RAF officer, Eric Jones, was brought in to assess the situation. His report, dated 2 February 1942 and classified Most Secret provides a fascinating and objective commentary on the work of Huts 6 and 3 in early 1942.17

  Jones’s conclusions and recommendations were very clear. He recommended the removal of Curtis whom he described as ‘a charmingly naive plagiarist who puts to the War Office as his own, interpretations borrowed from others’. He also recommended the removal of Humphreys, as he was ‘inclined to make personal capital out of the work’.

  The outcome of Jones’s report was the removal of Saunders from Hut 3 along with Humphreys and Curtis. They were initially replaced by a management triumvirate of the three senior officers in the Hut, but Jones himself was confirmed as overall head in July 1942. Jones would go on to replace Travis as Director of GCHQ in 1952.

  Welchman had been one of the first to recognize the indivisibility of the whole Enigma effort, from interception right through to intelligence. All sections – interception, cryptographic, traffic analysis and intelligence – operated as parts of a whole. Hut 6 also defined some basic principles of cryptography which would prove invaluable to Britain’s own intelligence security. An assessment of how the Germans measured up against these principles was included in Hut 6’s end of war report:

  1. The cipher must be theoretically secure – i.e. unbreakable if no errors are made in the construction of keys and encoding of messages. This was easily achieved by the Germans. While the first two indicating systems used by them were inadequate, the system introduced in May 1940 made Enigma theoretically unbreakable

  2. The construction of keys must be wholly random i.e. there must be no rules of keys or key repeats (other than such repeats as can occur by chance). The Germans never properly realized the importance of this and failed, particularly on Air Force keys.

  3. The effect of possible errors by the cipher clerks must be obviated. This was the hardest to achieve and the Germans failed despite great efforts and a wealth of ingenuity. The main reason why the theoretical security of Enigma did not give practical security was that the Germans failed to control their cipher clerks. They introduced new devices in a piecemeal fashion such as the plugable reflector D and the use of dummy words and repeated characters. The supervision of their cipher clerks was not thorough enough or sufficiently co-ordinated. German cryptographers were never allowed to inspect genuine traffic or keys.18

  By October 1943, Welchman’s main work for Hut 6 was complete. He had put in place an effective infrastructure, efficient processes, the provision of the tools to do the job and a high calibre of staff. Although improvements were still possible, he was needed for another role at BP. According to Milner-Barry:

  Welchman’s originality of mind, strong mechanical bias, and imaginative vision were invaluable assets for one presiding at the birth of an infant organization and planning its future growth.

  Chapter 4

  Turing, the Bombe and the Diagonal Board

  Three men, Alan Turing, Gordon Welchman and Harold Keen, were ultimately responsible for the machine which enabled BP to read millions of encrypted messages during the Second World War. Only Welchman would live long enough to see some information about their magnificent achievement reach the public domain. The first accounts were laughable and, in some cases, wildly inaccurate. Winterbotham was the first to attempt to describe a machine that he knew little about in his book The Ultra Secret in 1974:

  I am not of the computer age nor do I attempt to understand them, but early in 1940 I was ushered with great solemnity into the shrine where stood a bronze-coloured column surmounted by a larger circular bronze-coloured face, like some Eastern Goddess who was destined to become the oracle of Bletchley, at least when she felt like it. She was an awesome piece of magic.

  This was followed by Cave Brown in his book Bodyguard of Lies in 1975:

  The machine was installed at Hut 3, a large Nissen hut under the trees in Bletchley’s parkland, and the time soon came to begin operational trials by feeding Enigma intercepts to ‘The Bomb’. These intercepts were simply obtained from the string of tall-pyloned wireless interception posts which the British government had established around the world. The posts recorded all enemy, hostile and suspect wireless traffic and radioed it to Bletchley Park, where Enigma transmissions were identified, put on tape and fed into the ‘Bomb’. If the ‘Bomb’ could find the keys in which the transmissions had been ciphered, the cryptanalysts at Bletchley could then ‘unbutton’ these messages.

  Cave Brown had clearly been doing his research but unfortunately, had conflated three key developments at BP: a machine to help with the decryption of Enigma messages, the world’s first electronic computer which used paper tape to input data and was nothing to do with Enigma, and a hand technique developed by Dilly Knox to tackle messages encrypted with early versions of the Enigma machine. The first machine was the bombe, the second Colossus and Knox’s technique was called ‘buttoning-up’. Ironically, Cave Brown’s book was approved by the American authorities before publication.

  It is unlikely that Winterbotham and Cave Brown were trying to deceive their readers but, rather, were writing in ignorance of the real facts. In a modern computer world it is all too easy to think that the bombe was indeed a computer which magically worked out the daily key settings that operators on a German communications network were using on their Enigma machines. The reality was far from that, as Welchman pointed out in a letter to Robin Denniston in February 1984:

  The technical point that must be made to cryptographers of today is that we had a manual method and later on a Bombe that enabled us to ignore the enormous number of alternative plugboard connections in the first stage of the attack on an Enigma key. This, indeed, was all that the Bombes did.

  The use of machines by the Poles to assist in the attack on the Enigma machine has been described extensively in the literature and in Chapter 3 of this book. Rejewski published his own accounts which appeared in English in the early 1980s.1 Twinn confirmed to Welchman in their correspondence in the 1980s that Knox’s team had descriptions of the Polish bomba available in the Cottage. Knox had also used the idea of a crib with his hand methods to great effect in the early part of the war. As part of his team, Turing would have had access to this information.

  Alan Turing’s work and life has been admirably covered by Andrew Hodges in his fine biography, Alan Turing: The Enigma. In 1935, Turing had gone to a Part III course on Foundations of Mathematics given by M. H. A. Newman. Something important had lodged in Turing’s mind: was there a definite method, or as Newman put it a mechanical process, which could be applied to a mathematical statement, and which would come up with an answer as to whether it was provable? Max Newman would eventually be a colleague of Turing at BP and head the section named after him (the Newmanry) that would commission Colossus, the world’s first electronic computer. Stimulated by Newman, Turing went on to publish in 1936 his seminal paper with the tongue-twisting title of ‘On Computable Numbers, with an application to the Entscheidungsproblem’. It was a remarkable piece of work, summarised by Hodges as follows:

  Alan had proved that there was no ‘miraculous machine’ that could solve all mathematical problems, but in the process he had d
iscovered something almost equally miraculous, the idea of a universal machine that could take over the work of any machine. And he had argued that anything performed by a human computer could be done by a machine. So there could be a single machine which, by reading the descriptions of other machines placed upon its ‘tape’, could perform the equivalent of human mental activity. A single machine, to replace the human computer! An electric brain!

  Three years later, Turing found himself at BP, working in The Cottage with Dilly Knox. The eccentric ways of both gentlemen have been well documented and it is fascinating to imagine how these two characters got on, considering Knox’s apparent dislike of most of his male colleagues. Given Welchman’s difficulties with Knox, one wonders if Turing ever experienced anything of a similar nature. What does survive is a memo from Knox to Denniston in which the author says:

  Turing is very difficult to anchor down. He is very clever but quite irresponsible and throws up suggestions of all sorts of merit. I have just, but only just enough authority and ability to keep his ideas in some sort of order and discipline. But he is very nice about it all.2

  With Welchman dispatched to Elmers School to work on traffic analysis, it was Turing who began to look at how cribs and mechanization could help with the decryption process. A crib was a guessed phrase that could be matched character by character, to a string of encrypted characters in the message. A crib was independent of the German indicator system. Even before the Pyry conference, the idea of mechanizing two Enigma wheels three positions apart to exploit the repeated characters (known as ‘females’) caused by the German Enigma operators encrypting the message setting twice, had been discussed at BP.3 It is important to note that at this point Turing’s idea for a machine solution had nothing to do with his earlier ideas about a ‘universal machine’. What he envisaged was a bespoke solution, specific to the internal wiring of the Enigma wheels and constructed around the crib.

  As Turing’s ideas developed, Welchman had moved from Knox’s team to begin building his own team in Hut 6 and putting in place his ideas for expansion. Like Turing, he was also thinking about machine solutions for the Enigma-related problems facing them. With their ideas taking shape, BP needed to turn them into reality and for this they looked to BTM and the company’s brilliant engineer Harold Keen.4

  In 1910, BTM had been awarded the contract to mechanize the 1911 British census. To meet this commitment, the company had recruited new staff and so Keen joined in 1912. In 1921, most of the company, including Keen, moved to a new factory in Letchworth, Hertfordshire. By 1923, he was head of the Experimental Department and eventually became regarded as one of the most successful British innovators in the field of punched-card machinery. He had also picked up the name ‘Doc’ because of his habit of carrying papers and tools in a small bag, much like the one used by doctors. In 1939, Travis visited BTM and following discussions with Keen and the Managing Director, a Mr Bailey, a contract was signed to produce the bombe machines under the codename ‘Cantab’. Peter Twinn briefed Keen on the technical details.

  As the machines were urgently needed Keen adapted, where possible, suitable and proven components already being manufactured on a large scale for BTM’s existing products. The design, manufacture and assembly of the machine took place in his own department to ensure security. All BTM equipment was based on the requirement to record numerical and textural information by means of accurately punched holes in Hollerith cards. The holes were sensed by small metal brushes and used to activate electro-mechanisms. Timed electrical clock pulses were distributed within or between mechanisms by using commutators with circular rows of contact segments and rotating brushes. This technology, along with the company’s expertise in the use of electromagnetic relays as switches, lay at the heart of the machine that BTM would produce for BP.

  Keen decided to use circular wheels (known as drums) to replicate the wheels of an Enigma machine. Each so-called scrambler in the bombes (also called Letchworth Enigmas) consisted of three drums which were internally wired to replicate the Enigma wheels that they represented. Rather than sitting side by side like the wheels in an Enigma machine, the bombe’s drums were attached to the machine in vertical columns. Operators had to have clear access to individual drums by so that they could easily change them during use. Each drum had a different colour to help the operators to identify them and the colours used for wheels I–VIII were red, purple, green, yellow, brown, blue, black and grey respectively. Only the German Navy used wheels VI, VII and VIII.

  In early versions the sensing rate of the commutator segments was 20 to 25 per second. Based on the lower rate it took 22 minutes to test all positions of three wheels and therefore to test the possible wheel configurations available to the German Army and Air Force operators took 22 hours. This was eventually reduced to around 15 hours. Standard BTM units and components were employed to provide the various functions.

  Twinn had told Keen that up to 40 scramblers (each with three drums) might be needed to satisfy Turing’s design requirements. A compromise was agreed that limited the number in the first two prototypes to 10. The next two prototypes incorporated a significant modification invented by Welchman (the Diagonal Board) and had 12 scramblers. By the end of 1941, the bombes had 3 banks of 12 scramblers, close to Twinn’s original prediction. To provide the flexibility to interconnect the various components of the machine, the machine had 230 × 26-way jacks and batches of 26-way plug-in cable connectors.

  By the summer of 1941, there were four to six bombes at BP in Hut 11 and a similar number in some converted stables at Adstock Manor in the village of Adstock, where coincidentally, Travis lived throughout the war. The machines were delivered in ordinary unobtrusive covered lorries like the ones used for BTM’s own machines and therefore attracted no undue attention. Eventually some machines were installed in the villages of Wavendon and Gayhurst, 70 in all. Then a purpose-built facility was opened in Stanmore in August 1942 and in 1943 at Eastcote. By the end of the war, BTM had produced around 211 machines.

  The Enigma machine had a weakness in its design and in its use. Its weakness in design was that it could not encrypt a letter as itself. Its weakness in use was that the operators, particularly those in the German Air Force, became lazy and over-confident and their continual habit of reusing the same phrases in messages day after day, provided Hut 6 with an ample supply of cribs. The standard procedure was to write the letters of the guessed phrase and the intercepted message in its encrypted form on slips of paper. By placing them on a table, one above the other, one slip could be moved left or right. What the cryptanalyst was looking for was a position where each letter in the phrase had been encrypted as one of the other 25 letters in the alphabet. The relationship between the guessed phrase (called a crib) and corresponding encrypted characters in the message could then be tested on a bombe. If the test was successful, part of the setting of the Enigma machine that had produced it would be revealed and further hand testing might reveal the rest of the setting. A detailed description of cribs and how they were used at BP can be found in Appendix 2.

  A bombe did not magically tell BP’s cryptanalysts how Enigma operators on a German communication network were setting up their machines before they encrypted a message. It simply eliminated enough of the 158.9 million, million, million possible ways they could have done it to allow the actual settings to be worked out using hand methods. John Herivel, who worked in Hut 6, wrote the following elegant description of the bombe many years later, using the German word ‘stecker’ for plug:

  The whole basis of Turing’s approach was the idea of running a crib through all 17,576 different positions of the drums for a given wheel order and accepting or rejecting a given position by testing to see if possible stecker pairs could or could not be found for it. This was a very remarkable idea as although in theory the steckers made the identification vastly more difficult, in practice they provided a method for spotting possible positions. Thus the steckers were used to work for a solu
tion and their ‘difficulty’ was thus neatly compensated for by their ‘utility’.

  There were problems with Turing’s original design as only certain cribs worked effectively. The solution came to Welchman in flash of inspiration and involved interconnecting the scramblers in the bombe in a way that meant that almost any crib could be effective. As he wrote in The Hut Six Story:

  When this new method of interconnecting the scramblers came to me, I couldn’t believe it. But I sat down with a few coloured pencils, drew a simple wiring diagram, and convinced myself that the idea would indeed work. Armed with this diagram I hurried once again to the Cottage, this time to talk to Turing. On this occasion I had a better reception than I had received from Dilly. Turing was incredulous at first, as I had been, but when he had studied my diagram he agreed that the idea would work, and became as excited about it as I was. He agreed that the improvement over the type of the Bombe that he had been considering was spectacular.

  Welchman took the idea to Travis and with Turing’s backing he was dispatched to brief Keen at BTM. His idea proved to be reasonably straightforward to implement and Keen grasped the principle behind it quite quickly. It became known as the diagonal board and speeded up the solution of Enigma keys by orders of magnitude. Welchman and Keen worked closely together for the next few years and also become close friends. Bombes with a diagonal board became the default configuration for the machine for the rest of the war. Welchman had wanted to call it the modified device the ‘spider’ to distinguish it from the early prototype but, in the end, the name bombe was used for the rest of the war. However, Travis tended to use the term spider when referring to a bombe.