In October 1942 Travis and Captain Carl F. Holden, the US Navy’s Director of Communications, concluded the wide-ranging ‘Holden Agreement’ (sometimes referred to as the Travis–Wenger Agreement) on naval Sigint. Under the Agreement, work against Shark became the subject of ‘full collaboration’ between GC&CS and OP-20-G. GC&CS was to help the US Navy to develop ‘analytical machinery’ (the bombes) and there was to be a complete exchange of technical information on Kriegsmarine ciphers. The Agreement was the first to establish the vital Sigint relationship between the two countries during the war.
The British official history errs in claiming that ‘in September [1942] the Navy Department announced that it had developed a more advanced machine [i.e. a bombe] of its own, [and] would have built 360 copies of it by the end of the year’, and that the Agreement included a ‘compromise’ under which the ‘Navy Department… undertook to construct only 100 Bombes’. No US Navy bombe of any kind had been developed by September 1942. Neither the original US Navy directive for the bombe project nor the Holden Agreement specified the number of bombes to be built. OP-20-G’s decision to build only about ninety-five bombes was not taken until March 1943.
One of the first two US Navy bombe prototypes, named Adam (the other was ‘Eve’), was delivered to the Navy Computing Laboratory in Dayton on 26 May 1943 to begin tests. There were countless problems, with both machines, especially with ‘shorts’, ‘opens’ and oil leaks. Neither machine was running well as late as July 1943. However, the problems were eventually sorted out, and production models began to be shipped to Washington, DC, in late August. Some production bombes were in ‘semi-continuous operation’ by 11 August, with six operating by 7 September.
The standard US Navy bombe, known as a ‘530’, was about 2 ft wide, 8 ft long and 7 ft high and weighed 5,000 pounds. The British three-rotor bombe was a few inches smaller, but weighed only about 2,200 pounds. The Navy bombe had at least three hundred valves – quite a large number for that time. The Navy’s 530 type bombes only had sixteen banks of Enigmas – eight on each side (see page 7 of the plate section). They could therefore only run short menus, with a maximum of sixteen letters. OP-20-GM eventually built three double unit bombes, known as the 800 bombe (nicknamed ‘granddad’), which contained thirty-two banks of Enigmas, to tackle ciphers such as Seahorse. The first 800 unit became operational at the end of January 1944.
About ninety-five of the Navy 530 type bombes were made, including eight inverted machines. A 530 bombe ran a three-rotor menu in a mere fifty seconds, and a four-rotor menu in twenty minutes. However, due to the time required to set up a bombe menu and change rotors between runs, the average production per watch of eight hours was between forty-eight and sixty three-rotor runs (known as short runs) or twelve four-rotor runs. Twenty-three units of an improved model, the 1530, with circuitry which reduced the number of false ‘stops’, were also built. The US Navy also used a number of ‘black boxes’, called ‘grenades’ (being small bombes), which were attached to bombes to deal with various Enigma problems where the Stecker were known. The standard grenade, for example, was used to locate the rotor starting positions when the rotor order, ring settings and Stecker had been found.
Once the Navy bombes were moved to Washington, they worked well, and were extremely reliable. Thus in April 1944, when eighty-seven Navy bombes were in service, their down-time was about 2.7 per cent, while routine maintenance required about 2.5 per cent. This was just as well, since the British four-rotor bombes were far from reliable. In early 1944, only twenty-five or so four-rotor bombes had been installed at GC&CS. In March, their performance was described by Hugh Alexander as ‘still poor, and likely to remain so’. Very few, perhaps only three, were actually operational then. In consequence, GC&CS decided that future British bombe production would be concentrated on making three-rotor bombes, although a further eighteen or so four-rotor bombes were in fact made.
The US Navy bombe figures for maintenance and reliability were maintained consistently throughout 1944 and 1945, each being under 3 per cent. Like some British bombes, some US Navy bombes continued to operate after the end of the war in Europe: one Navy bombe was still attacking old wartime ciphers in March 1946. It has even been claimed that some were brought out of storage to tackle an East German police version of Enigma in the early 1950s.
After Hut 8 re-entered Shark in mid-December 1942, OP-20-G broke a few keys manually. By the end of 1943, with about seventy-five bombes in service and considerable experience under its belt, OP-20-GM was allocated responsibility for breaking Shark. At first, OP-20-G used cribs sent by Hut 8. However, they soon started to devise their own cribs. By August 1944, Hut 8 had only four cryptanalysts (Rolf Noskwith, Joan Clarke (one of Bletchley’s very few female cryptanalysts), Patrick Mahon and Richard Pendered) – very few, bearing in mind that a three-shift system operated, and that there had been sixteen in February 1942. The number of Shark keys solved by Hut 8 gradually diminished. It solved only five Shark keys in January 1945, none in February and one in March.
The US Navy bombes very soon had spare capacity. Although the Navy was very apprehensive about trespassing on the US Army’s preserves, it agreed to run Hut 6 (Army and Air Force) problems (codenamed Bovril). However, tackling this work for Hut 6 ran counter to an allocation agreement with the Army on the division of work that had been agreed by President Roosevelt in 1942. Eventually, a somewhat reluctant Travis was pressed by Rear Admiral Joseph Redman to obtain the approval of Colonel Carter Clarke, the chief of the Army’s Special Branch, for the work being done by the Navy. Clarke agreed, but gently chided Travis and Redman for having put the work ‘into operation without our prior knowledge’. In January 1944, about 45 per cent of the Navy’s seventy-five bombes were on Hut 6 work, while about 60 per cent of its 115 bombes were so engaged in January 1945. The number of Navy bombes on Hut 6 jobs therefore rose from roughly about thirty to seventy during 1944.
GC&CS had about seventy three-rotor bombes in service in January 1944, and 140 in January 1945. Allowing for set-up times, and changing rotors, one US Navy four-rotor bombe could do about three times the work of a GC&CS three-rotor bombe. The seventy Navy bombes on Hut 6 work in January 1945 were therefore equivalent to about 210 GC&CS three-rotor bombes, increasing GC&CS’s three-rotor bombe capacity by 150 per cent. This was extremely fortunate, since GC&CS needed all the bombes it could get. Indeed, in February 1944, GC&CS had asked OP-20-G to order fifty additional bombes. They were duly requisitioned, and started to enter service in mid-August. However, the existing US Navy bombes were found to be so efficient that the additional order was reduced to twenty-five in September.
From June 1943 to June 1944, OP-20-GM recovered:
236 Shark keys (65 per cent of the total)
266 special Shark keys (65 per cent of the total) – these were probably keys modified by the use of a special cue-word (Stichwort) procedure
488 other Kriegsmarine keys
581 German Army and Air Force keys.
The Kriegsmarine began to issue special Enigma ciphers (Sonderschlüssel) to individual U-boats in mid-1944, but did not use them until mid-November. Sonderschlüssel were virtually unbreakable, since very few cribs were available: it took 5,300 bombe hours and about six weeks’ work before Sonderschlüssel 161 succumbed in early April 1945, and only three Sonderschlüssel were ever broken, all by OP-20-G. The widespread use of Sonderschlüssel from February 1945 onwards deprived the Allies of virtually all operational intelligence on the U-boats. However, even worse was to come. On 1 April 1945 Plaice, a cipher used in the Baltic, implemented ‘one of the most formidable changes’ made by the Kriegsmarine, by employing a set of 288 Grundstellungen for April, instead of a single daily Grund. Hut 8 had to break a considerable number of individual messages in order to reconstruct the list of Grundstellungen, and could do so only because a new machine, Filibuster, was available. Hut 8’s problems were compounded when Dolphin adopted the new system on 1 May – and a fresh set of 288 G
rundstellungen took effect for Plaice. Hugh Alexander was by no means certain that Hut 8 could have survived these changes. At least twenty-four more examples of Filibuster and another machine (Hypo) would have been required, as well as a greatly increased staff. Even then, the outlook would have been uncertain.
Anglo-American co-operation on naval Enigma proved extremely successful, due to the excellent relations between Hut 8 and OP-20-GM. Without the help of OP-20-G’s superb four-rotor bombes, Shark and the other M4 ciphers could not have been consistently broken after mid-1943. Hut 8 broke about 1,120,000 out of the 1,550,000 Kriegsmarine Enigma signals intercepted during the war, although only 530,000 decrypts were sent to the Admiralty, since many dealt with weather forecasts and other matters which did not affect operations, even indirectly. It was a magnificent performance, given that Hut 8’s complement never exceeded about 150 staff. Naval Ultra was the most important of the many intelligence sources, including photo-reconnaissance and the invaluable HF-DF fixes, available to naval intelligence during the war. Without Ultra from Shark, in particular, the course of the war would have been very different, since it must be very doubtful whether the Allies could have established naval supremacy in the Atlantic until the second half of 1943. Ultra was only one of many factors, including shipbuilding capacity, modern sensors and weapons, trained seamen and airmen, which contributed to the Allied victory, but Hut 8’s crowning achievement was to save countless lives, on both sides of the conflict, by helping to shorten the war.
12
HUT 8 FROM THE INSIDE
ROLF NOSKWITH
Introduction
Chapter 12 gives an insider’s views on breaking naval Enigma. Rolf Noskwith was a young Cambridge mathematics graduate when he was recruited for Bletchley under a new system headed by C. P. Snow.
He joined Hut 8 in June 1941, when it was starting to get into its stride, and was assigned to its crib room because of his fluent German. Hut 4 (Naval Section) had attempted to provide cribs until Hut 8 set up its own crib room in 1941, even though there was very little plain-text in 1940 from past decrypts, and virtually no continuity in breaking the few Enigma daily keys that had been solved. Frank Birch had erroneously believed that if Alan Turing, in Hut 8, tried out cribs with the bombes ‘systematically they would work’. But Birch had misunderstood the problems involved in using cribs, largely because Turing was ‘a lamentable explainer’. Some of Birch’s suggestions would have involved testing tens of thousands of results for a single crib, which would have been completely beyond Hut 8’s scant resources in 1940 and the capacity of the sole bombe that was available until towards the end of the year.
However, as Hut 4 gained experience and more decrypts became available, Hut 8 came to appreciate the very close co-operation that developed between its crib room and Naval Section, including the latter’s intelligence sub-sections, a unit breaking hand ciphers and the U-boat plotting room there. When a second black-out threatened to develop in March 1943 on Shark, the cipher used by the North Atlantic U-boats, sterling work in Naval Section helped to save the day, and to prevent the black-out from lasting for the several months originally forecast.
Hut 8 would probably not have solved the Porpoise naval cipher when it did if Hut 4 had not kept pressing it to investigate the traffic, which Hut 8 thought used a much more complex system than emerged. Fortunately, the liaison between Huts 4 and 8 on all topics became much closer and more successful as the war progressed, although surprisingly it was not until May 1943 that regular meetings began between Hut 8 and the relevant parts of Naval Section. Happily, at the end of the war in Europe, Birch was able to write to the head of Hut 8, Patrick Mahon, of ‘two independent entities so closely, continuously, and cordially united as our two Sections’. Co-operation, whether with other Huts at Bletchley, with OP-20-G, or with GC&CS’s outposts, was the key to the many superb Allied successes in the codebreaking war.
RE
At the outbreak of the Second World War I was an undergraduate at Cambridge, reading mathematics. As I also spoke German I thought that I had some qualifications for ‘decoding’ work. This view was accepted by the University Recruiting Board and included in a wider recommendation for my war service. There was then a setback when I failed two medicals. I therefore continued with my studies and in 1940 I was interviewed for work at what I later recognized to be Bletchley Park – but my appointment was vetoed because of my foreign birth. In 1941 I had a similar interview with C. P. Snow (the future novelist who was then a Civil Service Commissioner) and Hugh Alexander. The rules had since changed. 1 was appointed at a salary of £250 per annum as a Junior Assistant (later Junior Administrative Officer) at the Foreign Office which had jurisdiction over GC&CS.
I arrived at Bletchley on 20 June 1941, the day after my twenty-second birthday. Alexander met me at the station and took me through the Park entrance to one of the Nissen-type huts which had been erected in the grounds of the Victorian mansion. This was Hut 8, the unit of which Alexander was acting head and for which he had recruited me. It was only then that I was told that our task was to break the ciphers used by the German Navy. Despite my talk about decoding I had no conception of the work I would be doing and I entered a completely new world when I learned that the German Navy enciphered its signals using a machine called Enigma, which was also used by the German Army and Air Force.
I joined the crib section of Hut 8, headed by Shaun Wylie. We were known as cribsters and our function was to produce cribs, a crib being a guess of what a portion of a particular signal might be saying. A correct crib, tested on the bombes (the electro-mechanical machines developed by Alan Turing), would lead to the solution of a day’s keys.
Next door to us was a section, staffed mostly by mathematicians, which used a sophisticated application of probability theory, designed by Turing to reduce the number of variables which the bombes had to test. The process required the manipulation of big sheets of paper, each with punched holes representing an individual signal. These sheets were manufactured in Banbury: the process was therefore called Banburismus and the people doing the work were known as Banburists. There were ways in which cribsters and Banburists could help each other.
When I arrived in Hut 8, the keys for June and July 1941 had been captured by ‘pinches’, i.e. successful raids on German naval units, mostly small ships stationed in the remote North Atlantic to supply weather reports. Until 1 August we were therefore reading Enigma messages almost as quickly as the Germans for whom they were intended. While valuable intelligence was obtained from this material we studied it carefully in order to identify potential cribs.
A significant proportion of the messages turned out to be ‘dummies’, consisting of nonsense words like DONAU DAMPFSCHIFF FAHRTS GESELLSCHAFTS KAPITAEN (Danube steamship navigation company captain), further text such as ‘intended to deceive the enemy’ and finally a jumble of letters. Their purpose was to maintain a fairly even flow of traffic so that no inferences would be drawn from an upsurge in the number of signals when some special activity was planned.
The most common cribs were derived from weather reports. The Germans sent out regular messages, usually twice a day, from weather stations at various ports, e.g. Boulogne, Hook of Holland, Royan etc. A typical message would read ‘from weather station Boulogne x weather forecast for’, followed by the area and period of time covered by the forecast. The operator sending out such a signal had strict orders to vary the text from message to message, but not uncommonly these orders were disregarded so that the same text appeared day after day. At other times it was possible to discern a shift system, three operators on a rota, each with his pet text which then occurred at predictable intervals.
Even then there could be numerous variations in the wording of a weather report. It was possible to narrow the choice by making use of an important characteristic of Enigma: a letter in the plain-text could be encrypted as any other letter but never as itself. For example, if a signal had ‘V as its first letter, the underly
ing message could not start with ‘von’, the German for ‘from’. If we had a crib for the message starting with VON, we would say that it ‘crashed’. Conversely, if we had a long crib, say twenty-five or more letters, which did not crash, there would be an increased chance that it might be correct.
The captured material ran out at the end of July so that 1 August 1941 brought a great challenge for Hut 8. By an apparent stroke of amazing good fortune the day produced two probable weather messages (identified by frequency, code sign and time) which had both been encrypted with the same starting position of the wheels in the machine. Their beginnings were:
1. MON.O 2. MVV.E
It seemed highly probable that the decrypts would begin as follows:
M O N . O .
V V V (W) E .
M V V . E .
V O N V O N
This was a perfect example of a ‘depth crib’, combining the probable plain-text of two or more signals, where the starting positions were identical or in close proximity. The letters MV in the first column illustrated a ‘click’, where a common letter in both signals had to have a common decrypt. The subsequent parts OV/VO, NV/VN, OE/EO were known as ‘reciprocoggers’, i.e. paired letters which might be encrypts/decrypts of each other: such pairs increased the probability that the solution was correct. VVV and VONVON were alternative forms used in signals to mean ‘from’, and the messages were likely to continue with the German for weather station, i.e. WETTERWARTE or, more frequently, WEWA. This led to a small number of permutations which were almost certain to include a crib of sufficient length for the bombes. Despite this hopeful start we failed to achieve a break, and 1 and 2 August remained among the very few days which were never read.
The Bletchley Park Codebreakers Page 23