However, M4 was not a true four-rotor machine, since beta was the right-hand half of a split reflector. Being thinner than rotors I to VIII, beta was not interchangeable with them. Beta could be set (although it did not rotate during encipherment), giving M4 the equivalent of twenty-six different reflectors (in conjunction with thin reflector B), but M4’s rotors could still only be permuted in 336 (8×7×6×1) different ways – not 3,024 (9×8×7×6). Given cribs, three-rotor bombes could therefore attack Shark in M4 form. However, Hut 8 needed twenty-six times as many bombes to do so effectively, but held only about twenty three-rotor bombes in February 1942, and thirty in August. Few could be spared from their work on other keys to mount parallel attacks on standard M4 Shark signals. In March 1942, a proposal to release bombes from other work to attack Shark had to be referred to the Y Board for decision, which required the Chief of the Imperial General Staff, among others, to be briefed on the issue.
Soaring shipping losses on the Atlantic convoy routes from the autumn of 1942 onwards led the Admiralty to become somewhat disheartened. On 22 November 1942, the OIC urged Hut 8 to focus ‘a little more attention’ on Shark, thereby demonstrating that it had completely failed to understand the difficulties confronting Hut 8, which was helpless without cribs. Fortunately, Hut 8 was almost ready to defeat Shark.
On 30 October 1942, the second edition of the Wetterkurzschlüssel was seized by Lieutenant Anthony Fasson, Able Seaman Colin Grazier and a sixteen-year-old canteen assistant, Tommy Brown (all from HMS Petard), from U-559 before it sank north-east of Port Said. Tragically, Fasson and Grazier were drowned when the U-boat went down suddenly. They were awarded posthumous George Crosses, while Tommy Brown, who survived, became the youngest holder of the George Medal. The Wetterkurzschlüssel reached Hut 8 on 24 November. By 2 December, Hut 8 believed that it could soon re-enter Shark through the weather short signals. In order to ensure compatibility with existing machines, M4 emulated M3 at one setting of beta, which was Shark’s undoing. Moreover, Germet 3 additive tables, consisting of several sets of ten tables, were repeated, although in a different order, during the second of a pair of months. When this happened, Germet 3 could be broken with few delays. But in December Hut 8 had to wait for the 100 reciphering tables used in November with Germet 3 to begin to be repeated on 8 December before it could confirm its hunch, and for the tables to be ‘hotted up’ by being broken more fully. Hut 8 then found, after numerous bombe runs, that M4 was in M3 emulation mode when enciphering the weather short signals. A full three-rotor bombe run against M4 in M3 mode, using a menu derived from the weather signals, took about four bombe-days, instead of the 100 or so required when M4 used its full potential. It was easy to test the setting for the fourth rotor, and the key for a second day in a pair could generally be solved by three-rotor bombes in under twenty-four hours.
On 13 December 1942, Hut 4 sent a teleprint to the OIC, listing the positions of over twelve Atlantic U-boats, as established from Shark weather signals for 5 to 7 December. It was the end of the Shark black-out: eighty-eight out of the following ninety-nine days’ traffic on Shark were solved with the aid of the weather short signals and the DAN weather broadcasts broken by Hut 10, although sometimes there were significant delays.
Hut 8’s use of the Wetterkurzschlüssel against Shark only lasted for about three months. On 8 March 1943 Commodore E. G. N. Rushbrooke, the Director of Naval Intelligence, pessimistically advised the Admiralty that GC&CS would be blinded against Shark for ‘some considerable period, perhaps extending to months’, presumably after learning that from Bletchley Park. On 10 March, a third edition of the weather short signal book came into operation, again depriving Hut 8 of Shark cribs. Hut 8 reconstructed the substantially revised edition of the Wetterkurzschlüssel quickly, but could not build the monthly indicator tables used with it quickly enough to make much use of the signals. However, by using cribs based on U-boat short signal reports about convoys (which were encoded with a codebook known as the Kurzsignalheft), Hut 8 broke into Shark again on 19 March. Hut 8 solved Shark for 90 out of the 112 days from 10 March to 30 June, mostly with cribs from the sighting short signals. The sighting reports also used M4 in M3 mode – and a copy of the Kurzsignalheft had been captured from U-559.
The short sighting signals seldom provided cribs longer than seven letters, but they could be combined with other data to form bombe menus. Using the signals as cribs stretched Allied HF-DF and other intelligence-gathering resources to the limit. In Hut 4, a section under Edgar Jackson helped to assemble all the available evidence. Jackson’s mastery of the mass of raw material proved invaluable, especially since garbles, mistakes by intercept operators and unusual short signals laid many false trails.
A severe drop in the Shark traffic in the summer months, following the withdrawal of the U-boats from the North Atlantic convoy routes, coupled with the adoption by M4 of a new ‘Greek’ rotor, gamma (with thin reflector Cäsar), on 1 July 1943, led to serious delays in breaking Shark until September. British and US Navy production four-rotor bombes entered service in June and August 1943 respectively, but some July and August keys still took up to twenty-six days to solve. However, from September on, Shark was often broken within twenty-four hours (especially on the second day of a pair), although there were some difficulties and delays, especially when cribs were in short supply, as in October 1944, when the average delay in solving it was forty-five hours.
Hut 8’s successes owed much to the drive and energy of Hugh Alexander, who became its head around November 1942, when Alan Turing was in America. He had been the acting head for some time, since Turing was, to put it mildly, uninterested in administration. Alexander had moved to Hut 8 in March 1941, as deputy head under Turing, when naval Enigma started to take off after the Lofoten captures. Alexander was both a superbly skilled cryptanalyst and an excellent manager, who transformed Hut 8 into an outstanding machine for delivering decrypts speedily to Hut 4.
Alexander was keenly interested in staff morale. He applied ‘need to know’ in a positive way by telling the Hut 8 clerical staff, most of whom were doing monotonous (but vital) jobs, as much as possible about the operational results of their work, in order to help them appreciate that they were a crucial part of the machine. This worked well, and his trust was fully rewarded. Interestingly, Richard Feynman adopted exactly the same approach with some junior staff while working on the atom bomb at Los Alamos. The Hut 8 cryptanalysts were constantly in contact with the clerical staff, so that it was always clear to them that their work was being used, and not just being filed away. Unlike Alexander’s counterparts in Hut 6, he did not employ university graduates on clerical work, since he found that it led to dissatisfaction.
Alexander seems to have acquired organizational skills during the short period when he was in business before the war. Under his leadership Hut 8 constantly refined its techniques for dealing with various tasks. In particular, Banburismus was broken down into as many separate modules as possible. Charts prepared for the necessary calculations transformed a highly sophisticated Bayesian system into one in which the ground work could largely be carried out by unskilled clerks. When Banburismus began in mid-1941, ‘scoring’ tetras (repeated groups of four letters or more) in decibans (a unit of calculation invented by Turing) was too time-consuming even for skilled cryptanalysts. Later, a series of tables enabled two clerks to score over 100 tetras in a shift quite easily merely by looking up the tables, with the only mathematics required being simple addition and subtraction. Decibans played a vital part in Banburismus. Jack Good, another member of Hut 8, found that by changing the unit to half-decibans whole numbers could be used, instead of decimals. The change saved an immense amount of time and, more importantly, enabled Hut 8 ‘to succeed where it would otherwise have failed’.
Initially, Hut 8 consisted of two subsections: an inner room with the cryptanalysts, and a ‘Big Room’ with the supporting clerical staff who prepared work for the Banburists, kept traffic records, and tested th
e results from the bombes. As the work grew, the subsections were subdivided. There was first a ‘Banburists’ room’ and a crib room, then a research section was added and later a research crib room was created. A separate registration room was set up when traffic began to be teleprinted from the Scarborough intercept centre in April 1941. When bombes became plentiful, a separate bombe testing room was formed. The machine room decoding was at first done by Hut 6; it was later dealt with part-time by Big Room typists, and eventually a separate decoding room was established.
The solution of a cipher known as Porpoise in late summer 1942 meant that the registration room had to deal with a further 250 messages each day, and that the clerical staff were hard pressed. However, the main problems with Dolphin and Porpoise had been solved – breaking them was largely a matter of refining the ‘production line’ dealing with them – and the purely cryptanalytic work eased off. Three cryptanalysts (Harry Golombek, Leslie Yoxall and Perkins) were therefore transferred to other sections in October and November 1942. Alexander believed in keeping staff busy, not only to keep them interested, but also because he thought that work tended to be done poorly if they were underemployed.
Hut 8 broke many other ciphers besides Dolphin and Shark. However, one of its eventual successes was later a source of considerable embarrassment to it, being an episode over which Hut 8 ‘would gladly draw a veil of considerable opacity’. About 100 messages a day of Mediterranean traffic had been sent to Naval Section by bag during 1941 and early 1942 but, since the traffic did not use the Dolphin bigram tables, Hut 8 did not examine it closely. By 1942, 200 to 300 signals were being intercepted each day, which led Hut 8 to study the traffic for a few days in the early summer. It found that if the second letters of the indicators in two or more messages agreed, so did the sixth; the same applied to the third and seventh and the fourth and eighth letters (e.g. with XFZT IPAM and OFGS EPQS, F ‘throws-on’ to P in the same place in each indicator). Even so, Hut 8 thought that bigram tables were involved until a visitor from Hut 6 pointed out that it was a ‘throw-on’ system, such as was used by army and air force traffic before 15 September 1938, and in naval Enigma before 1 May 1937, with the Grund being set out in the key-lists, and not chosen by the operator (see Appendix II).
Throw-on systems were relatively easy to break given 100 or more messages a day. It was even possible to do so given between ten and fifteen messages. Moreover, the system was the only form of service Enigma that could be solved entirely without cribs, since it could be broken using only the enciphered indicators. Hut 8 quickly developed systems for breaking the cipher, whose German codename was Süd (codenamed Porpoise by Bletchley). Since no K book was involved, the operator was permitted to choose his own message key, subject to a few very basic rules.
While the German blunder is astonishing, it is also surprising that Hut 8 did not spot the nature of the indicating system, since when describing the weaknesses of the throw-on system in ‘Prof’s Book’ Alan Turing had written:
This phenomenon [of patterns in the indicators] enables us to tell very quickly with any cipher whether the Boxing form of indication is being used.
Hut 8’s failure to spot the nature of the system earlier also indicates that Turing himself had been taking comparatively little interest in the day-to-day work of Hut 8 from at least early 1942 onwards. He would almost certainly have recognized the nature of the indicating system immediately if he had studied the indicators for even a few minutes.
Süd eventually split into three separate ciphers: Grampus (German codename, Poseidon), Trumpeter (Uranus) and Porpoise (Hermes, formerly Süd). All used doubly enciphered message keys, as did three further ciphers. Porpoise and Grampus did not abandon the throw-on system until 1 June 1944. Sunfish (whose German codename was Tibet), a cipher used by the blockade runners returning from Japan, was broken by Hut 8 in August 1943, largely at the instigation of OP-20-GM (the US Navy’s naval Enigma section), which had made detailed studies of the traffic. Hut 8 had thought that Sunfish used a form of unsteckered Enigma, but it turned out to be another throw-on cipher. However, the US Navy bombes could not run menus derived from throw-on indicators, since they had only sixteen banks of Enigmas, and throw-on menus required about thirty-two. OP-20-GM therefore asked Hut 8 to run a menu on the Bletchley bombes, which contained thirty-six banks. All throw-on menus were expensive in bombe time, since the bombes stopped much more often than with ordinary menus. Hut 8 therefore attacked the traffic only when a blockade runner was returning home, and largely left Sunfish to OP-20-GM, especially after January 1944, when they were able to use their ‘800’ double bombes to attack the throw-on systems.
Hut 8 then broke Seahorse (whose German codename was Bertok, derived from ‘Ber[lin]-Tok[io]’) in September 1943, again at the request of OP-20-GM, following an extensive examination of the traffic. Seahorse was employed for communications with the German naval attaché in Tokyo, Vice-Admiral Paul Wenneker, and was called ‘Kriegsmarine’ by OP-20-GM. It, too, used a throw-on indicator system, and turned out to be an M4 cipher.
The final throw-on system attacked by Hut 8 was Bonito (German codename, Eichendorff), which was first broken, ‘[a]fter a certain amount of trouble’, in May 1944. Bonito provided intelligence on midget submarines and saboteurs. The traffic increased from five to ten messages daily in April 1944 to 100 a day in spring 1945.
The use of doubly enciphered message keys was a catastrophic blunder by the Kriegsmarine. Such a basic security error is almost incomprehensible, especially since the Heer and the Luftwaffe had abandoned double-encipherment on 1 May 1940 because they had become aware that it made Enigma very vulnerable. The Kriegsmarine was definitely fully informed about that decision, but still adopted a completely flawed system. In consequence, Hut 8 and OP-20-GM penetrated no fewer than six ciphers, some of which they probably could not otherwise have broken, since no cribs were available for them. Süd and its subsets alone ultimately yielded over 115,000 decrypts, containing a considerable amount of intelligence about the areas they covered.
The traffic handled by Hut 8 increased considerably as the war went on. On average about 460 intercepts were dealt with each day in 1941 and 1942, 980 in 1943, 1,560 in 1944 and 1,790 in 1945. The principal naval intercept station, Scarborough, had about sixty-five receivers on German naval traffic in April 1942, most of them taking naval Enigma. There were also a number of sets abroad on naval work, including some in the Commonwealth, and a small station in Murmansk, in northern Russia. Traffic often had to be double banked, since bombe menus depended upon accurate message texts, preferably with no omissions; sometimes Hut 8 even requested treble or quadruple banking, particularly for keys in the Mediterranean, where intercept conditions were very poor. Hut 8 did not control the intercept sets, unlike Hut 6. Instead, it requested special coverage through Hut 4, but the different system worked well in practice.
In addition to being a significant source of intelligence, shore high-frequency direction-finding (HF-DF) was Hut 8’s vital ally, especially when using the weather and sighting short signals as cribs. Moreover, without shore HF-DF it is highly probable that Banburismus would have been impossible. The only weak point in the K book indicating system was its use of a single Grundstellung each day in any one cipher. The Kriegsmarine could easily have provided for different Grundstellungen, but doing so early in the war would probably have increased the length of signals, making them more vulnerable to shore HF-DF. There can be little doubt that the Kriegsmarine’s fear of shore HF-DF strongly influenced its decision to adopt a common Grundstellung procedure. Shore HF-DF therefore indirectly advanced the production of operational intelligence from Dolphin by about one year, since only Banburismus enabled Hut 8 to overcome the shortage of bombes for naval work before mid-1943. By September 1942, seventeen naval HF-DF stations had been established in the United Kingdom. By May 1943, there were about twenty-three British and Commonwealth stations on the Atlantic seaboard. The US Navy also had a substantial number of A
tlantic stations.
Other forms of Sigint, such as Tina (the identification of radio operators from the characteristics of their morse-sending) and radio finger-printing (RFP), which used cathode ray screens to identify individual radio transmitters, were of less assistance to Hut 8. So far as U-boats were concerned, RFP could seldom do more than distinguish between the transmitters used by the large Type XB and Type XIV boats, and those in the general mass of U-boats. Due to the high standard of German transmitters, RFP could not identify the transmitters of specific boats, which in 1944 led Naval Section to conclude that all attempts to do so should be abandoned as an aid to the production of Ultra from Enigma, although not for the purposes of general intelligence. Curiously, OP-20-G took the opposite view so far as using RFP and Tina combined to produce cribs for Shark was concerned, even though US Navy RFP gave worse results than in Britain. Intelligence derived by Hut 4 from traffic analysis – the study of the external characteristics of signals, such as their indicators (for example, ‘w w’ for weather short signals), the volume of traffic, the radio frequencies used and so on – also helped Hut 8 on occasion.
Hut 8 in the long run gained much from its eventual close co-operation with OP-20-G but the relationship took some time to come to fruition. Mrs Agnes Driscoll, a celebrated US Navy codebreaker, and a small section in OP-20-G had begun an attack on Dolphin in November 1940, but did not have the slightest success. Following various disputes with GC&CS during the course of 1942 over the role to be taken by OP-20-G in attacking Shark, Commander Joseph Wenger, the head of OP-20-G, became exasperated by GC&CS’s failure to build four-rotor bombes. On Wenger’s recommendation, Vice-Admiral Frederic J. Horne, the US Navy Vice Chief of Naval Operations, therefore decided on 4 September 1942 that OP-20-G should embark on an extensive bombe programme, tentatively costed at $2,000,000 (the eventual cost was $6,000,000, about $48,000 (£12,000) per bombe). By 17 September, Joseph Desch, of the National Cash Register Company in Dayton, Ohio, had completed a twelve-page report setting out an outline design of a four-rotor bombe for the US Navy.
The Bletchley Park Codebreakers Page 22