The Bletchley Park Codebreakers (25 page)

The impact was most dramatic in the Italian section. Before the Sinkov mission, the section was reading two Italian diplomatic systems, neither of which produced significant intelligence. At Bletchley, Sinkov acquired information (including reconstructed codebooks) concerning two high-grade diplomatic systems hitherto unknown to the Americans, as well as details concerning a new version of TR, one of the systems already under study at the SIS. This information so accelerated the work against Rome's communications that in the Munitions Building six analysts were added to the four already committed to the Italian problem. By the summer of 1941 they were decrypting messages in IMPERO, the first high-grade Italian cipher read by the SIS, and the number of decrypted Italian messages deemed worthy of circulation began to increase.

Some of the material contributed by GC&CS (e.g. Red Army codes) could not be immediately exploited by the Americans because the SIS lacked the necessary staff or could not intercept traffic. Other material further illuminated difficult problems even if it did not provide a key to solutions. Although GC&CS had not cracked high-grade German diplomatic ciphers, the British had determined the nature of these systems. They confirmed for the Americans that the Reich foreign ministry used four different systems and that all used the same codebook (DESAB), each differing only in the encipherment applied to that code. At the SIS, Solomon Kullback's German section had reconstructed the codebook, but for lack of traffic was struggling with the enciphered version known as
Spalierverfahren
. The British also lacked sufficient traffic to make much progress against
Spalierverfahren
, but they provided information about the encipherment and actually donated a set of cipher tables recovered from the German consulate in Reykjavik during the British occupation
of Iceland. From his new friends at Bletchley Park, Kullback also learned that Berlin's high-grade diplomatic system enciphered DESAB with a one-time pad, while the next system in importance (dubbed Floradora by the British) enciphered with a long additive. GC&CS told the Americans that, for the moment, it had abandoned work on these formidable ciphers, but at least the Americans now knew what they were facing.

As both London and Washington intended, the Sinkov mission launched a programme of cryptanalytic exchanges. For its part, the SIS wasted little time in exploiting the new relationship. Sinkov and Rosen had hardly unpacked their bags before the SIS was asking GC&CS for information regarding Vichy French diplomatic ciphers, especially those used on the circuits connecting Vichy and French territories in the western hemisphere. In return, the SIS began sending to Bletchley Park cryptanalytic observations regarding German, Italian and Japanese systems. These contributions, while earnest, were not always relevant to the problem at hand. An early consignment of German material included American observations relating to ciphers used by Germany in the First World War.

In August 1941, Commander Alastair Denniston, the operational head of GC&CS, crossed the Atlantic to observe American cryptanalytic operations and discuss further collaboration. At the SIS (he also visited OP-20-G) Denniston was pleased to note that collaboration on German diplomatic ciphers was making progress towards solving a problem that had seemed intractable. He was less sanguine about the new Vichy French and Latin American sections that the SIS had opened with the assistance of material from GC&CS. Here, too, collaboration was promising, but Denniston hoped that the Americans would concentrate their resources on Japan and not be distracted by forays against secondary targets that the British could cover well on their own.

By the end of 1941, collaboration against diplomatic targets was sufficiently established that the Japanese attack on Pearl Harbor on 7 December had little impact on SIS-GC&CS relations beyond accelerating the pace of co-operation. For example, by March 1942, when Lieutenant-Colonel John Tiltman, chief of the military section at Bletchley Park, arrived in Washington for a four-week visit to American signals intelligence facilities, GC&CS had added to its list of items passed to the SIS all the Vichy French codes and ciphers
it was reading, as well as the basic codebook used by the Spanish foreign ministry and a description of its encipherment for high-grade traffic. Tiltman brought additional gifts. Instructed by Alastair Denniston to effect a ‘complete interchange of all technical knowledge available and in particular to hand over to [the Americans] all our technical documents', Tiltman arrived with a ‘considerable quantity' of material from the various sections at GC&CS, including three more Vichy codebooks, microfilms of Floradora material generated by GC&CS and ‘descriptions of the methods used for the solution of 3 or 4 different complex ciphers by our Research Section'. Tiltman also offered to produce upon request Brazilian and Portuguese codebooks.

In May 1942 the Americans further exploited British cryptanalytic experience when the Signal Intelligence Service despatched another mission to the Government Code and Cypher School. The American team, Major Solomon Kullback (senior SIS civilians had been given military ranks for the duration of the war) and Captain Harold Brown, spent most of their time at Bletchley Park, but they also visited the centre for diplomatic cryptanalysis recently established by GC&CS in Berkeley Street, London. At Berkeley Street, Kullback and Brown visited the sections working German, Japanese, Italian, French, Spanish, Chinese, Swedish, Near Eastern and Latin American diplomatic systems. Upon the mission's return to the United States in July, Kullback could assure his colleagues, ‘I found the British most helpful and co-operative … They were completely frank, open and aboveboard with me and kept no detail of their operation, procedures, techniques, or results from me.'

By the end of 1942 the Signal Intelligence Service (now ensconced in new headquarters at Arlington Hall, a former girls' school outside Washington) had significantly expanded the range of its diplomatic operations. Much of this expansion can, of course, be attributed to a dramatic increase in staff and resources after Pearl Harbor. The British connection, however, was a crucial factor behind the increasingly impressive performance of the SIS. Indeed, it would not be an exaggeration to say that in the period 1941–2, GC&CS fuelled American diplomatic cryptanalysis by generously sharing its experience and material. In many ways, the collaboration that emerged in that period was one-sided. With the exception of the Japanese problem (and that is an important exception), the American contribution to the common effort remained relatively modest throughout 1942, mainly
because the American cryptanalytic record against targets other than Japan was relatively modest. In the early stages of the joint effort against important targets – Germany, Italy, France, the neutrals – GC&CS simply had more to contribute.

It is perhaps indicative of the early history of co-operation that when Tiltman visited Washington in the spring of 1942 he brought a bag full of cryptanalytic gifts, but when he returned to Bletchley Park he arrived empty-handed. On the other hand, through the missions of Sinkov and Kullback, Denniston and Tiltman, GC&CS certainly obtained from its ally useful insights into particularly recalcitrant problems, such as the high-grade German diplomatic ciphers. On at least one occasion the British codebreakers frankly acknowledged that liaison with the Americans on the German systems, particularly Floradora, had been especially fruitful. Still, the British may be forgiven if they sometimes felt in this early period that they were contributing to the relationship more than they were receiving. Often the Americans were so unrevealing about their own work that GC&CS was not sure which foreign ciphers the SIS was reading. In the spring of 1942, for example, GC&CS complained that it required such information in order to determine what French, Italian, Spanish and Latin American material should be passed to the SIS. As late as December 1942 Alastair Denniston felt compelled to ask his liaison officer in Washington: ‘Strictly between ourselves, are the Americans making a massive library of foreign government systems for filing purposes, or do they actually work on the stuff which we send them? We hear so little about Spanish and even French results that sometimes we wonder if they are actually deeply interested.'

Despite such occasional frustrations, Anglo-American co-operation in diplomatic cryptanalysis was generally smooth. Indeed, fruitful wartime collaboration continued long after the Signal Intelligence Service (renamed first the Signal Security Service and then the Signal Security Agency) evolved into a large organization employing thousands of staff and directing a worldwide intercept network. As American diplomatic cryptanalysis matured, GC&CS continued to provide important support usually in response to direct requests for assistance. In early 1943, for example, Berkeley Street passed to Arlington Hall what British codebreakers knew about the wheel wirings of the Enigma machine used by the Swiss foreign ministry. In August of that year, Berkeley Street provided additional information
concerning the Swiss machine as well as full information on the Dutch Hagelin diplomatic cipher machine. In 1944, British contributions included full information (often together with reconstructed codebooks and cipher tables) concerning Bulgarian, Burmese, Greek, Hungarian, Iranian and Iraqi diplomatic systems. Such assistance was particularly welcome as Arlington Hall expanded its operations by opening Balkan and Middle Eastern sections.

Both parties were so satisfied with the state of co-operation that disagreements in other areas of signals intelligence collaboration, such as the famous ‘Enigma Crisis' of 1943, hardly rippled the waters of diplomatic operations. By that time co-operation against diplomatic targets was taken so much for granted that the SIS felt comfortable asking its operational sections to submit lists of what they needed from GC&CS in the full expectation that those needs would be satisfied. Further, the curious omission of any reference to diplomatic targets in the so-called BRUSA Agreement of 1943 – apportioning responsibilities for Axis military traffic – suggests that the informal exchange arrangements in the diplomatic area were so mutually satisfactory and non-controversial as to require no formal delimitation.

As might be expected, though, the collaboration was never perfect. Occasionally (especially towards the end of the war) one party would decline to share information. In the summer of 1944, for example, the Foreign Office hesitated to disclose to the Americans the contents of the so-called Reserved Series of intercepts, revealing foreign diplomats reporting the comments of British officials on various topics, including American policies. Later that year, the Foreign Office became increasingly skittish about Berkeley Street giving the Americans cryptanalytic information regarding certain countries, such as Egypt, that London considered clients. On one occasion GC&CS felt honour-bound to refuse an American request for a Polish codebook because the Poles had given the British a copy under a promise of security. For their part, the Americans were less inclined as the war wound down to share information about South American systems. Less frequently, one party would launch an operation without informing the other, as in February 1943, when the SIS created a highly secret section to study Russian ciphers. Ironically, the Americans remained unaware that in late 1944 GC&CS established its own secret unit to study Russian internal traffic (civil and military) in Sloane Square, London.
Of course, signals intelligence methods and successes have always been among the most precious secrets of any government. What is remarkable is not that London or Washington would hesitate or equivocate before exchanging such secrets, but that such hesitancy and equivocation were so rare.

By the end of the war, the United States had developed a large and productive signals intelligence organization that was reading the diplomatic traffic of almost every government in the world from Afghanistan to Yugoslavia. Much of this success was due to the diligence and skill of American cryptanalysts. No small part, however, was played by the British codebreakers at the Government Code and Cypher School. Throughout the war GC&CS provided timely advice and assistance that significantly advanced the American programme in diplomatic signals intelligence. The value of the British contribution was accurately summarized by the US Signal Security Agency in a postwar review of operations: ‘It is doubtful whether success in solution of certain diplomatic systems could have been achieved in time to be useful had not the British supplied the necessary information … The debt of the SSA to GC&CS in shortening the period between the beginning of study and the production of translations was in the case of the diplomatic traffic of certain governments very great indeed.' Without doubt, the British had been most helpful and co-operative.

11

BREAKING GERMAN NAVAL ENIGMA ON BOTH SIDES OF THE ATLANTIC

RALPH ERSKINE

Like all navies in the Second World War, the
Kriegsmarine
had to rely extensively on radio communications. In particular, U-boat pack tactics in the Atlantic were completely dependent on the use of radio. The Admiral Commanding U-boats –
Befehlshaber der U-Boote
(BdU) – directed the packs on the basis of sightings and other reports from the U-boats, which he collated with all the other intelligence available to him. To protect its signals, the
Kriegsmarine
employed the three-rotor Enigma machine, M3, with the three additional naval rotors, VI, VII and VIII, although it also used some manual ciphers, when security was not all important. At least twenty naval Enigma ciphers came into service during the war. The principal cipher in 1940 and 1941,
Heimisch
(later renamed Hydra, and called Dolphin by Bletchley Park), was used by U-boats and surface ships in home waters, including the Atlantic. Some naval ciphers, such as
Ausserheimisch
(Foreign Waters, later known as
Aegir
- codenamed Pike by Bletchley) were never broken during the war, although OP-20-G (the US Navy’s codebreaking unit) devoted much effort to attacking Pike in 1944. In addition to
Allgemein
(general) keys, most naval Enigma ciphers had doubly enciphered
Offizier
(Officer) keys, which were especially difficult to break. A few had
Stab
(Staff) keys, which Hut 8 (Bletchley Park’s naval Enigma section) broke on only one occasion.

Despite the fact that the Poles had given GC&CS a reconstructed Enigma and rotors I to V in August 1939, probably only two people at Bletchley Park believed in the autumn of 1939 that naval Enigma could be broken: Frank Birch, the head of the German naval subsection, and Alan Turing. Birch thought it could be broken because it had to be, even though Alastair Denniston, the operational head of Bletchley, had told him at the outbreak of the war that ‘all German codes were unbreakable’. Turing had become interested in naval Enigma when he arrived at Bletchley in September 1939, ‘because … I could have it to myself. Despite Denniston’s forebodings, Turing solved the complex naval indicating system (which is described in Appendix III) in December 1939, using some Polish decrypts for 1 to 8 May 1937. However, since he did not have the bigram tables used with the system, he could make no progress against wartime traffic. He therefore decided to tackle signals that had been intercepted in November 1938. After two weeks’ work, five days’ traffic was broken. It was found that six
Stecker
(plugboard connections) were still being used, and that no letter was steckered for two days in succession, which eliminated twelve letters on each second day. Turing hoped that this would allow him to make progress quickly, but he had made little headway by May 1940.

New wartime rotors, VI and VII, were recovered from the crew of U-33 in February 1940, but Dolphin proved much more resistant to
attack than the
Luftwaffe
Enigma ciphers. In May and June 1940, using papers captured from the patrol boat
Schiff
26, Hut 8 solved six days of April Dolphin traffic, with help from the first British bombe, ‘Victory’. Rotor VIII was captured in August 1940, after which Bletchley held all eight rotors. In November, after months of work, Hugh Foss broke the key for 8 May, which became known as ‘Foss’s day’, using the bombe. The key for 7 May was quickly broken, but only about three other daily keys were solved before February 1941.

None of the ten or so Dolphin daily keys solved in 1940 was broken in time to provide operationally useful intelligence. Hut 8 faced two main problems in trying to break Dolphin after the first bombe with Gordon Welchman’s diagonal board entered service in August 1940: first, it lacked cribs, which were needed on a daily basis to provide menus for the bombes. But none would be available until it had read a substantial amount of traffic, which it could not hope to break. Secondly, GC&CS had only two bombes at the end of 1940, which were seldom available for naval use. Birch complained bitterly that Hut 8 was not getting ‘fair does’, but to little avail. Without a short cut, M3’s eight rotors imposed an impossibly heavy load on the bombes, since up to 336 rotor combinations had to be tested, taking around eighty hours (excluding the time to change rotors) – almost six times longer than a comparable attack on the sixty rotor combinations involved in
Heer
or
Luftwaffe
Enigma. The resulting delay of four days or more would have rendered the decrypts useless for operational purposes. Moreover, at least two bombes would have been constantly required, and Bletchley could not spare them from other work for so long. Hut 8 therefore required some method of reducing the number of rotor orders to be checked.

Turing had in fact already invented a Bayesian probability process called Banburismus to solve the difficulty. Banburismus reduced the number of M3 rotor orders to be tested from 336 to a manageable number, usually between forty and sixty, by ascertaining which rotors were in M3’s right-hand and middle slots. However, to use Banburismus Hut 8 needed a complete set of the bigram tables for encoding Dolphin message indicators. Without the tables, Hut 8 had no prospect of aligning Dolphin signals ‘in depth’ by correctly superimposing cipher texts enciphered with the same key, which was the basis of Banburismus. However, Hut 8 lacked both the bigram tables and enough solved traffic to reconstruct them. To solve Dolphin
it required an extensive corpus of Dolphin plain traffic. This was a singularly vicious circle from which Hut 8 could escape only if the Royal Navy captured the tables or Enigma key-lists to allow Hut 8 to read the traffic in order to recreate the tables.

On 12 March 1941, Hut 8 received an Enigma key-list for February 1941, which had been snatched from the German armed trawler
Krebs
during a commando raid in the Lofoten Islands. Hut 8 had read most naval Enigma for February and had broken about eight days’ April traffic by 10 May. Although the decrypts were not available in time to be operationally useful to the Admiralty’s Operational Intelligence Centre (OIC), they enabled Hut 8 to reconstruct the ‘Bach’ bigram tables then in force.

Weather signals were transmitted daily by designated Atlantic U-boats, as an essential part of the German war effort, since so much of European weather moves from west to east. The signals were encoded on the
Wetterkurzschlüssel
, a short signal book, in order to shorten them as a precaution against shore high-frequency direction-finding (HF-DF), before they were enciphered on Enigma. The weather data were then rebroadcast by the powerful
Kriegsmarine
transmitter at Norddeich (call sign DAN), after being enciphered in the
Kriegsmarine
manual cipher called ‘Germet 3’ by Bletchley (and also known by it as ‘the DAN meteorological cipher’). Bletchley Park’s meteorological subsection in Hut 10 broke the DAN meteorological cipher from February 1941 onwards. On 10 May 1941, Hut 8 received the 1940 edition of the
Wetterkurzschlüssel
, which had been seized from the weather ship
München
in a specially planned operation. Hut 8 could now reconstruct the exact plain-text of the encoded weather short signals from the U-boats, giving it a further invaluable source for bombe menus.

Enigma key-lists for June and July seized from
München
and another weather ship,
Lauenburg
, proved a godsend. The resulting decrypts familiarized Hut 8 with naval Enigma sufficiently for it to solve the August Dolphin keys with an average delay of fifty hours. The lists also greatly eased the task of reconstructing a set of new bigram tables, codenamed ‘
Fluss
’, which came into force on 15 June. Hundreds of bombe runs would otherwise have been required to help build up the new tables. Since there were then only about six bombes. Hut 8 would probably have been unable to break naval Enigma currently for a further two months – until October – but for the key-lists.
The traffic also enabled Hut 8 to assess the extent to which the
Kriegsmarine
was sending dummy signals consisting of nonsense, in order to defeat traffic analysis: the signals would otherwise have falsified many of its language statistics, which were a vital part of Banburismus.

Despite its new-found knowledge, Hut 8 was unable to solve the Dolphin keys for 1 to 6 August and 18 and 19 September. However, those were the only days on which it failed to break Dolphin during the rest of the war. In finding naval keys, Hut 8 was helped because the inner settings (the rotor mix and ring settings) were changed only every two days, presumably because only officers were permitted to alter the inner settings, saving Bletchley a considerable amount of bombe time. In addition, since Hut 8 knew the rotor order for the second day, it did not have to carry out Banburismus, which was a time-consuming process, on the second day. This almost halved its work but, more importantly, it greatly speeded up the solution of Dolphin for second days. If a crib was available on the second day, a few bombe runs could therefore find the settings very quickly, since only a single rotor order had to be tested. After August 1941, Hut 8 broke most Dolphin traffic within thirty-six hours during the remainder of the war. Hut 8 took about three days to solve the settings for the first of a pair of days’ traffic in August, and under twenty-four hours for the second day. In October 1941, Dolphin took seventy-five hours to solve on the first day of a pair, but only a few hours on the second.

The June and July decrypts also enabled GC&CS’s Naval Section in Hut 4 to spot a breach of cipher security in the inter-connected German signal nets. Because some small
Kriegsmarine
units were not issued with Enigma, signals were sometimes enciphered on a manual system known as the
Werftschlüssel
(dockyard cipher), as well as Enigma. Hut 4 first penetrated the
Werftschlüssel
in mid-1940 and was breaking it more or less currently by March 1941. The deciphered
Werft
versions of the signals provided Hut 8 with a second source of cribs. On occasion, minelaying operations (known as ‘Gardening’) were carried out by the Royal Air Force to provide Hut 8 with
Werftschlüssel
cribs. The
Kriegsmarine
then sent signals in Dolphin and the
Werftschlüssel
about the re-opening of the relevant sea lanes after sweeping them for mines.