FIVE

Enigma

Cryptography, the art of secret communications, would be transformed by the advent of the cipher machine, particularly the Enigma, based by Arthur Scherbius on a secret writing machine invented by the Dutchman Hugo Alexander Koch. In 1923, Scherbius exhibited his machine with its striking art deco logo at the International Postal Union Congress and advertised it as a means of safeguarding the secrecy of cables and telegrams. What began life as a commercial enterprise was to end up as being a major influence on the outcome of battles. The businessmen of the world were slow in responding, however, but this was the year that British politicians had revealed how much wireless interception and Room 40’s codebreaking had contributed to the allied victory in the First World War. Germany had found means of defying the Versailles treaty, which only allowed the defeated enemy to take measures for security and counter-espionage, by calling its growing army the Reichswehr and its secret service the Abwehr, both implying defence; it now needed a means of hiding its secret communications. Scherbius seemed to be the answer as his pamphlet claimed that there were so many millions of ways of setting the machine that it would be impossible to break the messages without knowing the key settings. Germany seized the opportunity to acquire commercial machines and adapt them for military purposes.

GC&CS appointed Edward Travis, Alastair Denniston’s deputy, to take charge of all security aspects of British government communications and it therefore fell to him to investigate the commercial machine when the British patent was applied for in 1927. After an Enigma machine had been acquired for GC&CS, Travis asked Hugh Foss, who had joined GC&CS in 1924, to see if Scherbius’s claims that it was unbreakable were true. The elaborate sales pamphlet, produced in English under the title of The Glow-Lamp Ciphering and Deciphering Machine ‘Enigma’ showed that it looked like a typewriter in a wooden box with a standard keyboard (the continental QWERTZU instead of the British QWERTY) and above that a lampboard with a series of small light bulbs in the same order. The operator typed out the plain text message and the act of depressing the key sent an electrical impulse through the machine and the enciphered letter would show up on the lampboard, making commercial secrets safe for ever. Scherbius added in a non-committal way that ‘decisive battles have been lost, by land and by water, in the air and in debating with each other, because the adversary had a better method of keeping his correspondence secret’.

The British patent specification no. 267,472, accepted on 11 August 1927, set out exactly how the electrical ciphering machine worked. Stress was laid on the importance of being able to change the order of the three wheels, that it was possible to acquire extra, differently wired, wheels, and that ‘in the case of war’ there was the further advantage ‘that if surprised by the enemy it was only necessary to remove the wheels to render the machine useless’. The hint in the 1923 pamphlet had now turned into a bold statement about the importance of Enigma’s use in war. Accompanying photographs showed how to set the machine, the interchangeable order of the wheels, the clip or Ringstellung which set the moveable rings on the wheels and the Grundstellung, the basic starting position in which the letters on the wheels were to appear in the windows before enciphering the message key. The operator positioned the three wheels in a pre-determined order. The Germans chose to set the rings to the agreed Grundstellung and then selected three letters at random as the message setting. These were encrypted twice to produce a six-letter message indicator. The operator then set the wheels to the message setting and enciphered the message, pressing the key for every letter of plain text and reading off the enciphered letter on the lampboard. He sent the enciphered indicators in the preamble to his message. Foss painstakingly got down to the problem he had been set and was able to produce his report in 1928.

Foss called the machine the Reciprocal Enigma, as a previous large typewriter enciphering machine described in a patent of 1924 had only been a one-way machine with no reflector (Umkehrwalze) and consequently was non-reciprocal. The electric signal set off by depressing the key travelled through wiring of the wheels to the reflector, which sent the signal back through the wheels a second time and the enciphered letter lit up on the lampboard. The action of depressing the key also moved the first, or right-hand, wheel one notch forward. Periodically, the movement of this first wheel led the other wheels to move forward. This was known as a ‘turnover’. Foss considered that in spite of the high degree of security of the Scherbius Reciprocal Enigma, it could be broken, given certain conditions, if the codebreaker had obtained a piece of plain text or guessed it as a crib. If the wiring of the wheels was known, just fifteen letters were all that was needed to find the machine setting, but if the wheel wiring was unknown, at least 180 letters would be needed. It is uncertain whether, at this early stage, Dilly discussed the problem with Foss, who maintained that he was not making any attempt to provide a methodology for breaking an Enigma machine himself; that would be left to Dilly Knox to discover, but how much later is not clear. An unconfirmed statement quoted in Foss’s ‘Reminiscences on Enigma’, written in 1949, says that Dilly Knox purchased an Enigma machine in Vienna in 1925, which seems odd. With two young toddlers, he and Olive are unlikely to have been visiting Austria on holiday, although perhaps there was an exhibition of the machine which led to him going on his own to gather intelligence on how it worked. However, there is no doubt that there was one in GC&CS available after Foss completed his work in 1929; that being so, there is no question that Dilly would not have wanted to investigate it, whether or not, with all his other pressing Russian cipher commitments, he was officially asked to do so.

Scherbius had never given any suggestions about indicating to the recipient how the machine had to be set up to decipher a message, with or without a codebook of settings. When first analysing the commercial machine, without having any enciphered messages and with no indicators to work on, Dilly would have to investigate the machine’s characteristics and would soon discover that if the letter A was pressed and the light bulb showed K, in the same position K would light up as A; this reciprocal factor together with the fact that A could never produce A would be highly significant and would mean that initially he could approach the problem textually, as he was accustomed to doing. The first thing to do was to reproduce the action of the machine in a convenient manner to experiment on, as though it were a hand cipher; this he did by making lettered strips of cardboard, which he called rods, one for each wheel. He knew from testing the machine that what he called the ‘diagonal’, the wiring between the letter keys and the entry plate, was in the order of its keyboard, running from left to right, top to bottom, therefore beginning QWERTZU, the standard order of the German typewriter keyboard. This would figure largely in subsequent analysis of the machine.

Dilly could also observe from looking at the machine how the wheels turned over and deduced that it would be the right-hand one which, as the entry point for the current from the keyboard, would allow a stretch of up to twenty-five positions to work on the cipher text without a wheel turnover. Foss was quite specific that it was Dilly’s invention of rods that had achieved the break into the commercial Enigma machine. In ‘Reminiscences’, he admitted that ‘the methods I used were rather clumsy as they were geometrical rather than algebraical and, when Dilly Knox came to study the subject, he invented the “rods” and the process known as “buttoning-up”, which used the same properties as I had done but did so in a more effective way’. (For a detailed explanation of ‘buttoning-up’ see Appendix 3.) Dilly’s methods had nothing to do with algebra as such, of course; what Foss meant was that he had worked with diagrams and numbers and so missed out on all the linguistic potential that Dilly’s lettered rods would bring to machine cryptography and make them such an effective tool, when there were real messages to work on. (For a detailed explanation of ‘rodding’ see Appendix 2.)

The first sign of wars and rumours of wars when the Enigma machine might be used militarily was with Mussolini’s invasion of Abyssinia in 1935, which alarmed Britain and France and led to their co-operation in interception and other matters. Study of Italian naval traffic had already shown the growing interest of Italy in the Red Sea area and a jointly run wireless station was set up in southern France. Dilly joined ‘Nobby’ Clarke’s naval section to work with him on breaking the messages. The need was now urgent as Mussolini’s invasion threatened British-controlled Egypt and the route to India. In 1934, the Italian navy was still using super-enciphered codebooks and it was these that Dilly had to break to build up a background of naval vocabulary. The Italian naval attaché’s method was by additives, where the operator had to put a long line of random numbers under the code-groups and add the two together in non-carrying arithmetic. Dilly’s colleague Wilfred Bodsworth was impressed that Dilly was able to break the messages so effectively that, during the Abyssinian war, the Admiralty was kept fully informed of the strength and activities of the Italian navy.

The next threat imposed by the dictator Mussolini was even greater. He called the Mediterranean mare nostrum, as the Romans had done before him, and Gibraltar, Britain’s key to dominating access to the Mediterranean, was at risk. This was in 1936, during the Spanish Civil War, when Mussolini backed General Franco’s right-wing Nationalists against the left-wing Republicans. The Germans sold the Italians and Spanish a version of the commercial machine for their use and GC&CS could now receive the enciphered messages at the joint wireless intercept station in France. The first GC&CS break into Enigma would therefore be through Italian and not German traffic. The year 1936 was a watershed for Dilly as he was back into the excitement of Room 40 days with all the ‘interests which once our life comprised’ and he could now devote his time officially to breaking real operational Enigma messages, rather than ones he had concocted for himself in his spare time. It was that year that his family noted that something was afoot, as for the first time Dilly refused the invitation to the King’s College Founder’s Feast for fear of giving his new secret away over the port.

Dilly soon established that the Italian messages were not sent out on the commercial machine he knew; it was in fact an improved version of the commercial Enigma machine known as the ‘K’ model, in which the wheel turnover was attached to the alphabet ring and not to the wheel itself as in the ‘A’ model. The wirings of the wheels had been changed, making Dilly’s task more difficult, as Scherbius had recommended for increased security in wartime use, but now that he had genuine messages he could undertake a preliminary exercise which would hold good for any Enigma encipherment whatever the wheel wirings or message settings. This exploited the characteristic that A could not encipher as A and so if the Italian messages had a standard opening, say PER (Italian for ‘to)’, these letters would not appear in the first three places of the cipher text. Dilly undertook what he called a ‘boil’, which became a standard procedure for investigating new traffic. He took a piece of squared paper with numbers across the top and letters down the left-hand side and a dot was then put for the cipher text letters in the appropriate square; if indeed PER was a standard opening over a period of time this would show up on the chart as empty spaces in the first three squares. Unfortunately, there were not enough messages to draw any conclusion on this occasion and the next step was to hope for procedural errors by the operators which would give him the way in. Dilly’s work on Herodas meant he was used to slack Greek scribes and this approach was his forte in breaking Enigma messages.

Success came in April 1937 when twenty Italian naval messages were sent out on the same setting, contrary to strict training instructions given to operators; this meant that if the messages were written out under each other, each column of the resulting table would show plain text encipherment by a reciprocal simple substitution. It was an easy matter to start by letter frequencies and the guessing of probable words through repeated bigrams and trigrams or common opening words, provided the chosen letters did not ‘crash’ with the cipher text; that is to say that, since a letter cannot be enciphered as itself, if the message begins with the letter P, the message cannot have started with a word beginning with P. The guessed plain text was now written under the cipher text to produce ‘pairs’, each made up by the cipher letter followed by the plain text letter it represented as in the table below.

Cipher

R

T

F

B

R

X

E

Y

L

S

B

P

Y

S

P

Guess

P

E

R

X

C

O

M

A

N

D

A

N

T

E

X

Cipher

R

T

F

B

X

Q

C

R

L

Z

S

U

Guess

P

E

R

X

T

E

N

E

N

T

E

X

Cipher

P

S

R

F

C

Y

E

R

L

Z

P

U

Guess

R

I

F

E

R

I

M

E

N

T

O

X

Cipher

Y

F

N

F

C

X

G

H

M

O

X

X

S

Guess

N

U

M

E

R

O

X

C

I

N

Q

U

E

This produces the following resulting pairs in consecutive wheel positions: 1 = PR and YN; 2 = TE, SI and FU; 3 = FR and NM; 4 = BX and FE; 5 = CR and XT; 7 = XO, QE and VI; 8 = EM, CN and GX; 8 = YA, RE and HC. Using these pairs, Dilly could find the wheel wiring through the process he called ‘buttoning-up’, for which the rods, his outstanding contribution to machine cryptography, were made. Hugh Foss insisted that buttoning-up and the rods were Dilly’s unaided inventions. Dilly had to break each message by rodding as no attempt was made to rewire machine wheels to assist the task even though our own Typex cipher machines were being manufactured at the time.

It was soon found that Franco’s Enigma machine had the same wheel wirings, so that there could be communication with the Italians, which meant the same rods Dilly had made could be used for breaking Spanish messages. Wilfred Bodsworth, a Spanish expert, took over the task, employing Dilly’s methodology. In Spain, the Enigma machine was now being used for operational purposes and for the first time it would be seen how it performed in the field. Franco needed to keep in touch with his generals to co-ordinate military attacks in different parts of Spain, keeping the Republican units in the dark about his plans. Commander Antonio Sarmiento, who was in charge of training the operators, assured Franco that ‘to give you an idea of how secure these machines are, suffice to say that that the number of possible combinations is a remarkable 1,252,962,387,456’. He was unaware, however, in spite of his operator training, that this number could be considerably reduced through procedural errors, which were especially likely when a unit was on the move.

The Spanish Civil War gave Hitler insights into what modern warfare could now achieve, especially from the air. Although not officially backing Franco, under international law volunteers were allowed to give support in wars, as our volunteers to the International Red Brigade did to the Republicans. A volunteer unit of Luftwaffe pilots formed the ‘Condor Legion’ and were responsible for the bombing of Guernica, which alarmed the rest of Europe. They too used the type of commercial Enigma machine that Franco was using, which was broken thereby alerting Britain to Luftwaffe tactics and providing background knowledge for Dilly’s attempts to break the German air force Enigma. Dilly’s first break into an Enigma machine during the Spanish Civil War was an important cryptographical precedent for the Second World War, since it also ensured that, as war clouds gathered, the collection of intelligence from the deciphered messages was put on an operational basis.

Admiral Sinclair set up a committee to consider the implications of the Spanish Civil War, including the dreadful effects of mass bombing. Paymaster Lieutenant-Commander Norman Denning was appointed in 1937 to see what lessons were to be learned in linking cryptography and intelligence to best advantage. Mussolini had loaned four submarines manned by Italian crews to Franco to be used to blockade Republican ports and it was important for the Admiralty to establish the location of these ‘pirate’ submarines that were sinking merchant vessels and to keep track of any German or Italian ships in Spanish waters. Denning spent several weeks with Dilly and his colleagues following through interception and Italian and Spanish codebreaking and also made a detailed study of the First World War histories written by Birch and ‘Nobby’ Clarke in order to learn from Room 40’s experiences. He soon learned of the disastrous consequences of the separation of intelligence and operations, most notably at Jutland, and he began to plan and organise an Operational Intelligence Centre (OIC), where the two functions would be totally integrated. He got the principle of a new centre accepted and the OIC, although small, was up and running in 1939. Meanwhile, the new director of naval intelligence, Rear-Admiral John Godfrey, recruited suitable outsiders, forming a brilliant team for the Second World War and housing them in the Admiralty Citadel under Horse Guards Parade.

Another most important advantage gained by Dilly’s breaking of the Spanish and Italian Enigma was increased collaboration with the French Services de Renseignement, whose liveliest wire was the intelligence officer in charge of the cipher department, Gustave Bertrand. From 1926, when he first joined the service, Bertrand had received intelligence reports that the Germans had introduced an unbreakable Enigma cipher machine, first into their navy, then subsequently, in 1928, to the Reichswehr and, in 1934, to Göring’s newly formed Luftwaffe. The idea of Blitzkrieg, a war of total mobility between the armed forces co-ordinated using wireless communications operated from the backs of army trucks, was being worked out by the German generals. In May 1933, when Britain was still treating Russia as its chief threat, Bertrand had joined the crowds saluting the new Chancellor Hitler in his first appearance wearing the Nazi uniform with a swastika badge. Bertrand returned to Paris and warned that he had witnessed the rebirth of German militarism, but did not get much backing for increasing intelligence funding for the schemes he had been trying to put in place.

Although the French military had been receiving intercepted German messages from stations close to the frontier, they had not had any success in breaking them and Bertrand had decided that espionage was the only way forward. Fate played into his hands when in 1931 an official from the German military cipher office, Hans Thilo Schmidt, called at the French embassy in Berlin and astounded officials by offering to sell top secret documents to the French government relating to German organisation and procedure for enciphering their signals; his purpose was purely mercenary. Schmidt was given the codename Asché and Bertrand agreed to pay him 10,000 marks. The German handed over the documents, which included two operators’ manuals giving instructions on how to use the Engima machine and set the keys, at a rendezvous in a hotel in Verviers on the French–German border and Bertrand rushed upstairs, where there was a photographer waiting to copy them. Mission accomplished, Asché returned the precious documents to the secret safe from which they had come. Bertrand was greatly excited by the coup and bitterly disappointed when he was told by the head of the cipher department, a Colonel Bassières, that Asché’s documents would be of little use to the codebreakers without the wiring of the wheels or the key setting. ‘Impossible to get anything useful from your documents,’ he said, ‘too many things are lacking for us to produce the machine.’

A week later, Bertrand approached Wilfred ‘Biffy’ Dunderdale, the SIS station chief in Paris. Bertrand was the sole French liaison officer on intelligence matters and so had worked previously with Dunderdale, usually demanding payment for anything the French produced. Bertrand wanted a considerable sum for any more of Asché’s secrets so Dunderdale had to seek permission from SIS headquarters. The request was turned down flat. It was a political matter of funding priorities and it seems that Denniston, Foss, Tiltman and Dilly were not consulted. Dunderdale did have the original batch of documents for three days and in all probability photographed them, allowing Dilly to analyse them later, but the ban on paying any money for them cut the British off from the rest of Asché’s valuable secrets. Fortunately, when Bertrand turned to the Poles his reception was very different. The Nazi threat had become very disturbing for Poland, as there had been a tense relationship with Germany since the partitioning of Europe after the Versailles treaty. Hitler’s threat in Mein Kampf in 1926 that the Germans should find Lebensraum on their eastern border made it worse. The Poles were able to intercept German wireless traffic, which was transmitted at low power on medium frequencies, difficult to intercept in the UK, and they were particularly eager to know what the Nazis were saying. The Asché documents were received by the Poles as ‘manna in the desert’, according to Bertrand, and produced ‘an explosion of stupefaction and joy’.

One thing Dilly did learn at this time, although not having the advantage of the documents the Poles had received, was that the Germans had added a Stecker-board, similar to an old-fashioned telephone plugboard, to the commercial Enigma machine, which was clear from the photographs supplied by the French. This allowed many, many more different possibilities, thereby making the machine highly secure, if still vulnerable to the fact that no letter could ever be enciphered as itself. It would be another four years, however, before Dilly had any German service messages to work on and six years before he would finally receive Bertrand’s ‘manna in the desert’. The Polish codebreakers, meanwhile, had succeeded in breaking the German Enigma and were reading current messages by 1933, unbeknown to GC&CS. Bertrand was also ignorant of the Polish success, despite having contributed to it so much. Understandably, the Poles were anxious to conceal the break for security reasons, since if it reached German ears their ability to eavesdrop on their warlike intentions would have been lost. In retrospect, the early history of Enigma breaking and of appeasement might have been different if there had been co-operation between the wartime allies at this early stage. (The same security difficulty, concerning the secrecy over the breaking of Enigma, would later have to be weighed up by Winston Churchill in considering whether to share it with Americans before they came into the Second World War.)

Dilly first met up with the German Enigma messages during the Spanish Civil War, when their navy was on manoeuvres in Spanish waters; ominously by then they had renamed the navy Kriegsmarine, rather than retain its previous less offensive name, Reichsmarine. At this stage, Dilly was already aware of the introduction of the Stecker to the machine and, even before having any messages, he had worked out theoretically how they could be treated as a super-encipherment, which could be stripped off before applying his methodology for breaking un-Steckered machines. The Italians and the Spanish had used a separate codebook for the operators’ indicators for all the machine’s settings, but the Germans indicated the machine setting at the beginning of the message and this was for Dilly a new way in to codebreaking. Although the key setting of wheel order, Ringstellung, Stecker and Grundstellung was communicated in circulated German key lists, the operator chose his own setting, which he encoded on the basic Grundstellung and put the result at the beginning of the cipher text. If the indicator system could be unravelled it would mean that, when the machine was broken, a whole day’s messages on a given network might then be read, whereas with the Italian and Spanish codebook instructions, each message had to be broken separately.

Wilfred Bodsworth was now in charge of the Italian and Spanish traffic, leaving Dilly free to concentrate on the German naval indicating system as a means of breaking the machine. Alastair Denniston wrote that ‘Knox had made considerable progress in his diagnosis until April 1937’, but the navy then introduced a new procedural complication of bigram tables for the indicators and Dilly concentrated on unchanged army and air force traffic. John Tiltman remained in charge of the GC&CS Army section and a new air section to advise the Air Ministry, headed by Josh Cooper, had been set up in 1936. The discriminants (the groups in the message preambles denoting the particular network cipher system) of the new army/air force traffic between Germany and Spain were isolated by Tiltman’s deputy, Frederick Jacob, and its message indicators identified by Tiltman.

Hitler’s first belligerent move had been undertaken in March 1936 when over 32,000 soldiers and armed police entered the Rhineland, which had been demilitarised under the Versailles treaty. Although it transpired that Hitler would have withdrawn if France had taken action, nothing was done by Britain and France in response to his violation of the treaty, but eavesdropping was stepped up. This was the period when Tiltman made a breakthrough identifying what he described as ‘throw-on’ message indicators used by the German operators when, for reasons of clarity, they were told to tap out the setting of the three wheels on the fixed message setting twice and put it at the beginning of their message. Tiltman swiftly realised this meant that this would be likely to produce repeat letters in the key-block, the set of indicators at the start of the message, at positions 1 and 4, 2 and 5, or 3 and 6. If, for instance, there was an M in the first position, invariably there would be a certain letter, say G, in the fourth position, and so on; this proved that the indicators were on the same setting and ripe for exploitation.

In the autumn of 1938, Admiral Sinclair, anxious to increase co-operation with France, authorised Denniston to invite Bertrand over for a council of war. Bertrand, the man whom Denniston later called ‘a pedlar and purchaser of foreign government cribs’, arrived on 2 November 1938 and handed over a large number of documents, including original German material relating to Enigma that had been obtained by Asché. The French material was so important that Denniston wrote to Sinclair the same day asking permission to hand over a significant amount of different material in return. ‘We have received about 100 photographs of codes of which one might well have had great value in the event of war with Germany last month,’ Denniston told Sinclair. The French also handed over a full description of the Germans’ own communications intercept operations, some reconstructed German and Italian codebooks and a considerable amount of intercepted German and Italian telegrams. But the most prized documents were without doubt ‘photographs of documents relating to the use of the Enigma machine, which did increase our knowledge of the machine and have greatly aided our researches’, Denniston told Sinclair. ‘Our main reason for seeking this liaison in the first place was the desire to leave no stone unturned which might lead to a solution of the Enigma machine as used by the various German services. This is of vital importance for us and the French have furnished us with documents which have assisted us but we are still doubtful if success can be obtained without further documents.’ Sinclair immediately authorised the exchange.

The conference with Bertrand took place the following day and was held for security reasons at Sanctuary Buildings, Great Smith Street, a few hundred yards away from the Broadway headquarters. It was attended on the British side by Denniston himself, Oliver Strachey, as chief cryptographer, Tiltman and Dilly. The morning session began with a discussion of the various German armed forces Enigma material the French had produced, followed by a presentation by Tiltman on the British work on the indicators for the Enigma keys. After lunch at Simpsons in the Strand, Dilly explained the methods of breaking the latest commercial Enigma machines. The French were impressed. ‘It appears to us that their cryptographic work is less ambitious than ours,’ Denniston said. ‘They have worked on the German and Italian unreciphered codes with success and on the German military [double transposition] hand cipher.’

Bertrand was subsequently given the opportunity to see the flourishing Italian and Spanish section and Dilly’s methods of codebreaking and was promised a regular supply of decodes. ‘I think he was impressed by our success and by the ability of certain officers and it might well be that this had an important bearing on his subsequent action,’ Denniston later recalled. The whole Spanish situation was immediately apparent to the wily Bertrand, who understood that if Franco won his war against the Republican government and still used the same Enigma machine, France could keep an eye on its neighbour’s relationship with Germany, through its intercepts; bearing that in mind, Bertrand ‘salvaged five Spanish Republicans who had worked at Barcelona’ and installed them in Paris as intelligence officers in his new Spanish party.

There was an urgent request from the British for more material from the French agent and before the end of the year Bertrand handed over the German users’ manual for Enigma, which had been given to Rejewski in 1932. Schlüsselanleitung für die Chiffriermaschine Enigma, the directions for setting up the machine which, unbelievably, included a genuine example of a ninety-letter stream of text in both clear text and cipher, complete with key setting. This authentic example was soon replaced by a fictitious one in subsequent manuals, when someone in the German cipher authorities became aware of the blunder. The French material was passed over via ‘Biffy’ Dunderdale and arrived in GC&CS in the red-lined file covers in which SIS distributed its intelligence material, which we called ‘Scarlet Pimpernels’. If only someone had recognised the value of Bertrand’s version with its crib for GC&CS in 1932, Dilly might well have broken Enigma by 1938.

When Dilly finally received it six years later he immediately tried out his successful method for recovering wheel wiring, as he had done for the Italian and Spanish machines, but this time taking the German Stecker complication into account. Dilly had already decided that the Stecker merely added a substitution on the text letters entering the machine and the inverse substitution on the enciphered letters emerging. However, the way forward was not as easy as he thought. When Dilly was continually frustrated by not being able to recover the wirings by his well-tried crib methods, he feared that the diagonal, the entry connection from the keyboard to the wheels, was not QWERTZU, the typewriter order common to all the un-Steckered commercial machines he had worked on. QWERTZU was tried backwards to no avail and then came the awful thought that the Germans had achieved the ultimate Enigma machine complication – that of a random entry plate connection.

Bertrand set up a meeting with the Poles and the British in Paris in January 1939, which Denniston said was held in ‘an atmosphere of secrecy and mystery’. The Poles were very suspicious since the Munich appeasement with Hitler in September 1938 and still not prepared to reveal their Enigma successes. Denniston and the three codebreakers Dilly, Foss and Tiltman attended on the British side but the Polish representatives did not include any of their real codebreakers, with whom Dilly could have had a meaningful discussion. Hugh Foss recalled how annoyed Dilly was that ‘at these two interviews, the Poles were mainly silent but one of them gave a lengthy description in German of the recovery of “throw-on” indicators’. This was Major Maksymilian Ciężki of the Polish Cipher Bureau, who was in charge of the German section and had carried out the initial Polish work on the Enigma machine, before handing the material over to newly recruited mathematician codebreakers in 1932.

During Ciężki’s exposition at Paris, Dilly kept muttering to Denniston ‘but this is what Tiltman did’, while Denniston ‘hushed him’ and told him to listen politely. Thinking that Ciężki must know how the machine worked, Dilly naturally asked him the burning question as to what the diagonal he had assumed to be QWERTZU was. Dilly soon realised that the Polish officer knew nothing about the actual breaking of the Enigma machine and it was all a waste of time. He frustratedly wrote in his assessment of the Polish work on Enigma: ‘Practical knowledge of Qwertzu Enigma nil. Had succeeded in identifying indicators on precisely the methods always used here, but not till recently with success. He [Ciężki] was enormously pleased with his success and declaimed a pamphlet, which contained nothing new to us.’ It is clear that Dilly brought no information that would help him back from the Paris meeting with the Poles.

Dilly was also unimpressed by the French description of their attempts to solve Enigma, which they ended with a flourish and a dramatic ‘Voici la méthode française’. But he did strike up a lasting friendship with Bertrand’s chief cryptographer, Henri Braquenié, at the meeting and according to Foss, who was present at the time, the French were delighted when Dilly demonstrated how to use his rods and by the next meeting had made a set of reglettes of their own.

Denniston rightly realised that the Poles had agreed to the Paris meeting because they had run into difficulties in April 1938 when the Germans changed the indicating system. Colonel Gwido Langer, head of the Polish Cipher Bureau, had been told by Bertrand of Dilly’s progress in attempting to break Enigma and was anxious to seek his help. The minutes of the Paris meeting recorded frustration on both sides, Bertrand said, and ‘it seemed that the work had arrived at an impasse out of which only information from an agent could provide a way. A technical questionnaire was drawn up, as simple as possible, to give to the agent.’ Dilly asked Bertrand whether the agent could obtain sixteen alphabets enciphered on the Enigma machine for him, if he could not recover the wirings themselves, but any spy would be lucky to get away with half an hour’s noisy clonking of the wheels and the idea was abandoned. However, Dilly never lost hope that his methodology would finally succeed in spite of the pessimistic belief elsewhere in GC&CS that Scherbius’s claim that Enigma could be unbreakable if used properly was proving correct. ‘Nothing is impossible’ was after all Dilly’s motto.

Up to this time, Dilly’s only assistant on the German Enigma machine had been Tony Kendrick, but after the Paris meeting GC&CS decided to apply for a mathematician and appointed Peter Twinn from Brasenose College, Oxford. He recalled his first day at Broadway Buildings, where he found

the people working on the Enigma were the celebrated Dilly Knox and a chap called Tony Kendrick, quite a character, who was once head boy at Eton. There was a slightly bizarre interview with Dilly who was himself a bit of a character to put it mildly. He didn’t believe in wasting too much time in training his assistant, he gave me a five-minute talk and left me to get on with it.

Denniston now went further and made the rounds of the universities to bring in former colleagues and new talent, as Josh Cooper recalled:

He dined at several high tables in Oxford and Cambridge and came home with promises from a number of dons to attend a territorial training course. It would be hard to exaggerate the importance of this course for the future development of GC&CS. Not only had Denniston brought in scholars of the humanities, of the type of many of his own permanent staff, but he had also invited mathematicians of a somewhat different type who were especially attracted by the Enigma problem.

Alan Turing, of King’s College, Cambridge, went to one of the first of the training courses on codes and ciphers at Broadway Buildings; this was of a general nature and it is unlikely that Dilly was present, given his views on being thrown off at the deep end as far as breaking Enigma was concerned. However, Turing was put on Denniston’s ‘emergency list’ for call up in event of war and was invited to look in on Dilly’s small group to hear about progress with Enigma, which immediately interested him. Unusually, considering Denniston’s paranoia about secrecy, it is said that Turing was even allowed to take the ‘crib message’ back to King’s, and that ‘he sported his oak’, the then popular euphemism for working behind a closed door to discourage visitors, as well he might. They had begun to lose faith in Bertrand’s miraculous operators’ manual cipher and text crib and to think that what they were given to believe was an authentic example of cipher and plain text was after all fictitious. However, a meeting was arranged in Warsaw in July 1939, when the desperate Poles finally revealed, at the eleventh hour, how straightforward Dilly’s QWERTZU problem was and how near he had been to winning the Enigma race.