The Bletchley Park Codebreakers Read online

  Hut 6 solved Green for 25 October, the first wartime key to be broken in Britain, almost immediately after Turing’s return from France with the correct information about rotors IV and V. There was great relief when soon afterwards it broke Red (the main Luftwaffe cipher, which was widely employed for operational and administrative purposes) for 6 January, since there had been fears that the indicating system might have changed on 1 January. Hut 6 solved about fifty daily keys in Red, Green and Blue (a Luftwaffe practice cipher) between mid-January and late March 1940 using the Zygalski sheets. A new cipher, Yellow, was first intercepted on 10 April, during the Norwegian campaign, and was then broken each day until it went out of service on 14 May. Yellow carried both Luftwaffe and Heer traffic, and also gave some information about ship movements and German intentions. However, little use could be made of the decrypts operationally. Hut 3, which was responsible for analysing and collating the intelligence contained in the decrypts, as well as translating them, was formed into three eight-hour watches. At this stage its staff were mostly ‘amateur soldiers and airmen, unversed in the ways of military intelligence’, and lacking the necessary experience to deal with the plethora of codenames, abbreviations, arcane technical terms and other mysteries in the decrypts. There were not even enough staff or teleprinters to handle the flood of intelligence from Yellow and Red. To compound matters, the service departments had been expecting a silent war where radio silence ruled, and were insufficiently organized to handle the copious intelligence produced by Huts 3 and 6.

  On 1 May, the inevitable happened: all the ciphers except Yellow dropped doubly enciphered message keys. The Germans had at last realized that doubly enciphered message keys made Enigma vulnerable. All Enigma except Yellow became unreadable until 22 May, when Red for 20 May was solved using a highly ingenious method called ‘cillies’ (see Appendix IV), which had been discovered by Knox around late January 1940. Hut 6 abandoned work on other Enigma to solve about 1,000 messages on Red each day for the rest of May. The resulting flood of operational intelligence was teleprinted to Whitehall much more quickly than during the Norwegian campaign, and this time Whitehall was ready to forward it to the operational commands. Hut 6 depended completely upon cillies and related methods until the first bombe with a diagonal board entered service in August 1940, and continued to rely on them substantially for the remainder of 1940, and yet later whenever it had no other cribs. Red remained breakable with practically no gaps throughout the war, and was always the largest source by far of Hut 6 Ultra.

  At the July meeting with the Poles, Knox had realized immediately that the Polish methods were completely vulnerable to a change in the indicating system, since they depended upon doubly enciphered message keys. Knox and Turing therefore decided to develop a British bombe which would be immune to any such change. The first British bombe was ordered before 1 November 1939. It was completely different from the Polish model, since it used cribs to test certain assumptions about a key’s components, and did not merely check for ‘females’, which was the technique used by the Polish bombas. ‘Doc’ Harold Keen of the British Tabulating Machine Company at Letchworth was responsible for its detailed design, although the basic design concept, which defeated the all-important Stecker with their 150 million million key space, was the product of Turing’s genius. The bombes were not, despite some claims to the contrary, even remote forerunners of the computer, since their internal architecture bore not the slightest resemblance to that of a computer. Nor was Colossus (the Bletchley Park electronic computer) ever used against Enigma, as is sometimes suggested.

  The first bombe, named Victory, entered service on 14 March but it lacked the diagonal board designed by Gordon Welchman, and therefore produced far too many ‘stops’, which had to be tested by hand – a very time-consuming process. It was therefore only used against naval Enigma in a somewhat basic way, and not in attacking Luftwaffe or Heer Enigma. The first improved bombe (named Agnus, later corrupted to Agnes or Aggie), with a diagonal board, came into service in mid-August 1940. Bombes comprised thirty-six banks of high-speed electrically driven Enigmas (except for a few early models, which had thirty banks), with three or more diagonal boards. The diagonal board made it much easier to devise bombe menus, and reduced the number of false stops by 99 per cent – from 676 to four with an eleven-letter menu giving two ‘closures’. Depending on the composition of a menu, up to three rotor orders could be run on a single bombe. Setting up a bombe menu took thirty-five to fifty minutes, and changing a wheel order, ten minutes. A complete bombe run lasted about fifteen minutes, after the bombe was set up. Each bombe had two Wrens assigned to it: an operator and a tester, who made a preliminary check on the stop data to see whether they should be sent to a ‘testing party’ in Hut 6 for further examination. The two bombes in service in 1940 broke about 180 daily keys, all but a few (probably under ten) being Luftwaffe Enigma – mostly Red, which comprised about 120 keys. The daily keys (including naval Enigma keys) solved by the British bombes rose to a peak of 9,064 in 1943, and declined slightly to 8,444 in 1944, when US Navy bombes also solved a substantial number of keys.

  From about the end of 1940 onwards, the bombes were the essential basis of Hut 6’s solutions of Heer and Luftwaffe Enigma. Bombes at Bletchley Park were an inter-service resource, and were not ‘owned’ by any one Hut. Demands for bombe time by Hut 8 therefore had to be fitted in with the much more complex situation on Hut 6 ciphers. When Hut 8 suddenly needed a large number of machines, it could badly disturb the Hut 6 programme. Fortunately, very few issues of bombe priority had to be referred to Whitehall for guidance, since it would have been far too time-consuming a process: one example was at the end of March 1942 when the Y Board (which was responsible for overall Sigint policy) considered a proposal to allot six bombes out of about twenty then available to work on Shark (the four-rotor cipher used by the Atlantic U-boats) for about twenty-five days. In the event, only 7 per cent of the total bombe time was allocated to Shark in April – about forty-two bombe days out of 600. It was just as well, since the original proposal would have severely disrupted Hut 6’s work. Another major decision on priorities was taken in July 1942, when the PQ 17 convoy was under way and the fighting in North Africa had also become critical. It was decided that the threat to PQ 17 justified priority being given to the naval work on Dolphin. There were remarkably few disputes about sharing the bombes, partly due to the close friendship between Stuart Milner-Barry in Hut 6 and Hugh Alexander in Hut 8, and the fact that Milner-Barry fully appreciated the critical importance of the Battle of the Atlantic. A panel of five ‘bombe controllers’ – Milner-Barry, John Manisty and John Monroe from Hut 6, and Shaun Wylie and Hugh Alexander from Hut 8 – was established in mid-1942. The controller on duty settled all questions of bombe priority without leading to any friction between the two Huts.

  By May 1943, fifty-eight ‘standard’ three-rotor bombes, which did not print stop data, and fourteen ‘Jumbos’, which gave a printout, were being run by 900 Wrens and maintained by seventy-five RAF mechanics. By May 1945, 1,675 Wrens and about 265 men, virtually all from the RAF, were running and maintaining 211 bombes, including 66 four-rotor bombes for naval work. The bombe efficiency rate (the number of possible solutions, less those missed due to a bombe fault) hovered around 96 per cent throughout the war. Since each bombe contained about eleven miles of wire and 1,000,000 soldered contacts and received no preventative maintenance it was a minor miracle that they worked so well, twenty-four hours a day, seven days a week. A US Army contingent, the 6812th Signal Security Detachment, arrived in February 1944 to operate a bank of about eight bombes in the Eastcote bombe outstation from October onwards. The 6812th achieved a high degree of efficiency, with a daily average of 71.5 runs in April 1945 (twenty minutes per run, including the set-up time, etc.). It was a considerable accomplishment, since the theoretical maximum was seventy-two – the highest recorded figure by the Wrens was fifty-seven (twenty-six minutes per run).

  A good
crib for the bombes consisted of between twenty and thirty letters, but about fifteen would do at a pinch (for example, ‘Geraetklarmeldung’ - equipment ready). Cribs comprising as many as seventy-five letters were not uncommon, while occasionally they were 200 or more letters long. Considering that for all practical purposes cribs had to be letter perfect, it is amazing that so many long cribs were found. The following crib was part of a series called ‘Sultan’s Meldung’, which broke Red on no fewer than six separate days in April 1943:

  EINS X MELDUNG KLAM X LUDWIG KLAM YY … YY SULTAN X ROEM X EINS CAESAR X GEHEIM [One Report. (Ludwig) … Sultan Roman one Caesar [=Ic (the intelligence officer)]. Secret]

  It was easily recognizable, since it was transmitted daily on specific frequencies at three set times by the Luftwaffe’s Fliegerkorps X (codenamed ‘Sultan’).

  A frequent crib on Phoenix, which was used by Panzerarmee Afrika for its high-level operational communications between divisions and Corps, read:

  NAQT RUHIG YY KEINE BESONDEREN VORKOMMNISSE YY [Quiet night. Nothing special to report.]

  To Welchman’s regret, the general responsible for the unit sending it was captured, with the result that the crib died out.

  Hut 6 benefited greatly from the German reluctance to make the Stecker in Heer and Luftwaffe key-lists completely random. Stecker were not repeated on two successive days: if B and X were connected on Friday, that combination was not used on Saturday. The compilers of Luftwaffe key-lists also had a rule by which a letter was not connected to the next letter in the alphabet (B was not connected to C, C to D, and so on). Bletchley added a ‘Consecutive Stecker Knockout’ (CSKO) modification to the bombes to take account of this restriction, by eliminating stops which included consecutively steckered letters. A different rule prevented any rotor being used on consecutive days in the same position in the machine. Thus in the day before and after rotor order II, V, I was used, rotor II could not be used in the left-hand (slow) position, rotor V in the middle position or I in the fast (right-hand) position, reducing the rotor orders to be tested to thirty-two, instead of sixty. In addition, a rule under which no rotor order was used more than once a month was very helpful, especially towards the end of a month. And some ciphers started to use only half the available rotor orders, which became known as ‘Nigelian’ wheel orders (possibly being named after Nigel Forward, a member of Hut 6).

  Hut 6 also derived considerable help from the following basic mistakes by the German cipher operators:

  a) ‘Psillies’ – psychological cillies. In an indicator such as ROMXLV, XLV would probably be the enciphered message key ‘MEL’ (making ROMMEL when combined with the Grundstellung ‘ROM’), while TOBKST might be derived from TOBRUK.

  b) ‘Nearnesses’, which occurred when a cipher clerk, after selecting the Grundstellung (say HTB), calculated what the rotor positions would be three letters later (HTE here, if no rotor turnover was involved), typed them, producing, for example, DXX as the enciphered message key. Since his rotors were then at HTE, he did not have to reset them, and could therefore start typing the message text straightaway. The message key was then very ‘near’ to the Grundstellung – three letters away, so that this particular type of nearness was sometimes known as a ‘003’.

  The ROMMEL type of psilli and nearnesses each gave a three-letter crib, linking the unenciphered and enciphered message key.

  In September 1940, Hut 6 had broken Brown, which was used by the Luftwaffe’s KGr 100, a pathfinder type unit, and by the Sixth Company of the Luftwaffe’s Signals Experimental Regiment. Until Brown ceased in late 1941, it gave indirect warnings of many KGr 100 raids, as well as information which later enabled an advanced ‘beam’ radio bombing system known as the X-Gerät to be jammed. The Brown radio operators were so ‘peculiarly incautious in their radio chat’ that quite a lot of the intelligence attributed to Brown in fact came merely from ‘scarcely disguised plain language’ and not from decrypts. Daily keys could sometimes be solved by hand, especially on Brown, since it employed only six or seven pairs of Stecker instead of the usual ten. Moreover, Brown days were paired, with the letters linked by Stecker not being repeated on the second day of a pair.

  Heer Enigma was always much more difficult to break than Luftwaffe traffic, because Heer cipher discipline was always much higher than in the Luftwaffe, although the Luftwaffe improved in 1943. Heer Enigma included few cillies, or stereotyped beginnings or endings. Even in mid-1943, Dennis Babbage, the chief cryptanalyst in Hut 6, was worried that increased security precautions, such as the inclusion of ‘padding’ words at the start of messages and ‘burying’ addresses instead of leaving them at the start of signals, would soon prevent most Heer and Luftwaffe Enigma from being broken. Hut 6 wanted to find purely cryptanalytic or statistical methods to solve Enigma, but was never able to do so. Its worries and problems were to increase greatly as the war went on.

  In January 1942, Hut 6 was receiving about 1,400 intercepts per day (43,600 per month), including Railway Enigma traffic, and producing 580 decrypts each day (18,000 per month). By May 1943, the number of daily intercepts varied between 3,300 and 6,000 – between 102,000 and 186,000 per month. No figure for decrypts is available for May 1943, but it would probably have been about half of the intercepts. To handle so much traffic clearly required a very efficient and flexible organization. By May 1943 (the period to which the description in the next four paragraphs relates), Hut 6 had evolved into five main sections, with a total of about 400 staff: a central party known as No. VI Intelligence School (later renamed Sixta), control party, operations group, operational watch and research party (see Figure 4.2).

  The traffic intercepted by the Y service intercept stations was sent to Hut 6 by teleprinter so far as possible: the signals were preceded by a register with their preambles, which were transmitted immediately a full page was ready. Incoming traffic was first sorted by an identification party in the Hut 6 watch registration section into operational traffic, which was processed urgently, and non-operational, which had to wait. Operational messages then went to Registration Room No. 1, where a ‘discriminatrix’ sorted them by cipher system, using their discriminants and other data. Their preambles were then registered in a ‘blist’ (Banister list).

  Figure 4.2 Hut 6 organisation chart 17th May 1943, Head Gordon Welchman (400 staff)

  Operational messages were examined immediately by the cryptanalytic operational watch, which selected some from which to construct bombe menus. When a key was found by the bombes, the traffic using it was sent to the decoding room, where the signals were run through British Typex cipher machines, which had been converted to emulate Enigma. This part of deciphering the traffic was far from being a straightforward process: up to two-thirds of the messages could present deciphering difficulties – so-called ‘duds’ – messages which would not decipher, generally because they had the wrong indicators or contained garbled cipher text. The resulting decrypts were first routed through a room called the ‘cubicle’ for indexing, before being sent to Hut 3 for translation and analysis. A small section in the bombe control room (which was part of Hut 11A, where the Bletchley Park bombes were housed) processed duds. If the section could not decipher a message, it was sent to the log readers in the central party. By late 1944, the number of ‘tries and duds’ being attempted by Hut 6 had increased enormously, to an average of 1,125 per day (compared with 2,050 decrypts) in the week ending 7 October. Stuart Milner-Barry, the head of Hut 6 from September 1943 onwards, considered that a 2:1 ratio was about ‘the best that we can do under favourable circumstances’.

  The central party analysed the logs kept by the intercept operators in order to build up a geographical picture of the complex German radio nets, which constantly changed call signs and frequencies in order to outwit the British. Since Enigma cryptanalysis was inseparable from traffic analysis, the central party’s work made an indispensable contribution to Hut 6’s successes. By early 1943, it was a far cry from the days when GC&CS ‘did not think t
hat the results of traffic analysis were ever likely to help cryptanalysts’, and when Gordon Welchman was ‘deeply suspicious of Colonel Butler’s efforts to expand the W/TI [traffic analysis] organization’. John Coleman’s traffic research section in the control party was responsible for identifying the call signs used by the German transmitting stations under an intricate allocation system.

  The log-reading section in the central party was the biggest single unit in Hut 6, with ninety-five staff in May 1943. The log readers also looked for re-encipherments, routine cribs, cillies and ‘giveaways’ (cryptographic information in operators’ plain-language chat). Sometimes even the rotor orders or the Stecker were revealed by giveaways – operators on Brown were notorious for doing so. Re-encipherments occurred when a signal had to be sent from one cipher net, such as a Fliegerkorps, to another; they were more frequent at the beginning of a month, when new key-lists took effect. Re-encipherments could be identified from their times of origin, which were the same as the original signal, and their lengths, which were similar. The heads of the log-reading groups, with a few assistants, staffed the Fusion Room, which combined information from traffic analysis and cryptanalysis to build a complete picture of the Enigma radio nets, and fed it back in a unified form to the groups sending data to it.