There was nothing for the codebreakers to do but sit and wait, hoping that the Japanese had not picked up on the damaging leak and its attendant publicity. They did not have to wait long. In early August Nave's Royal Australian Navy intercept operators picked up Japanese messages telling stations to use an old, lower-grade code as emergency back-up on the orders of the ‘Head of War Inquiry Bureau’.
A week later, at 1500 hours Greenwich Mean Time on 14 August, two months into the new JN25 codebook, it was changed. Neither the American nor the British codebreakers were in any doubt as to the reason behind the sudden change, which was swiftly followed by changes to the other codebooks and a major change in the call-sign system. This combination of changes seems to indicate that the Japanese had read the US reports but that the inquiry into the affair concluded that the Allies had either worked out the details of the Japanese fleet from traffic analysis or somehow obtained a codebook and additive tables. The Japanese simply could not believe that anyone, least of all a Westerner, could break their codes.
A post-war Japanese history of wartime naval communications makes no mention of the US media reports, suggesting that somehow Japanese intelligence may have missed them, although it does confirm that they were worried that a codebook had been secured by the Allies. The Japanese history said:
With the failure of the Midway operation, it was feared that some of the codebooks lost were seized by the enemy. However, the staff of the Combined Fleet was of the opinion that if there were any codebooks lost, they would not cause an immediate danger. The Central Agency also held the same opinion and the only action taken was to apply the emergency regulations (using the old codebooks as the master code) as a temporary expedient. Subsequently, all the codebooks were revised in the order of their importance.
The change of the JN25 codebook led the British to protest a second time and to ask that no further public action be taken over the leak since ‘preservation of this invaluable weapon outweighs almost any other consideration’. Five days later, the Chicago Tribune published a front-page story claiming that it had been ‘cleared’ by the Grand Jury. The US Navy barred Seligman from all further promotion and there, finally, the matter was allowed to rest.
But as far as Washington was concerned, there was one matter left to resolve. Laurance Safford had been replaced as head of OP-20-G some months earlier by Commander John Redman, the brother of the Director of Naval Communications, Rear Admiral Joseph Redman. The Redman brothers were determined to ensure that communications experts were in charge of naval codebreaking. John Redman also appears to have resented being made to look foolish by Rochefort over the Midway disagreement. Word began to spread within the naval hierarchy in Washington that the naval intelligence centre in Hawaii was not working well and Rochefort was to blame. The esoteric codebreaker with a penchant for wearing a silk smoking jacket over his uniform was never likely to go down well with the tub-thumping Redman brothers. In an extraordinary memo that never once mentioned Rochefort by name, John Redman complained to his superiors that the Hawaii codebreaking operation was in the hands of a man who was merely ‘an ex-Japanese-language student’.
Rochefort was replaced and sent back to San Francisco where he was put in charge of the commissioning of a new dry dock. ‘What a waste of a priceless talent for a political payback,’ said Phillip Jacobsen. ‘Nimitz's recommendation for the Distinguished Service Medal for Rochefort was twice denied, but given to political cronies of the Redmans in Washington.’
12
FRIENDS FALL OUT
The US Navy codebreakers may have been enjoying some of their best moments of the war but their Royal Navy counterparts had reached their lowest point. The new recruits were finishing the Bedford course just in time. After undergoing codebreaking training, Jon Cohen was sent to the Japanese naval section in Elmers School at Bletchley Park with five other graduates of the first Tiltman course, including Hugh Denham, who had been just nineteen when he was recruited from Jesus College, Cambridge, where he had been reading Classics. The section now had around forty members. Hugh Foss was still in charge with Lieutenant-Commander Bruce Keith as his deputy and chief intelligence reporter, Hugh Denham recalled.
We were introduced all round and then went in to Keith for instructions. He said that there were about a thousand encrypted Japanese naval messages intercepted every day, of which some seven hundred were in JN25. A fraction of these reached Bletchley within days – in fact under forty intercepted at Flowerdown [near Winchester, in Hampshire]. Some of the remainder were in due course forwarded by bag from elsewhere. Current material was handled expeditiously at Washington, Kilindini and Melbourne. The Bletchley party had, up to that point, he said, achieved nothing. Whenever they began to attack a problem, the solution arrived from another centre and they turned to something else.
Keith told the new recruits gloomily that there was no work in the section for six Japanese linguists. The new boys were squeezed into a small room, recalled Jon Cohen. ‘At some stage a bomb fell on the building, purely an accidental target, as it were. We were sitting round the edge of this room and the whole ceiling fell down into the middle of the room. It was just fortunate that every one of us was sitting round the edges of this room and so we weren't hit by anything. We made some attempts to break the Japanese Merchant Navy Code JN40, though we didn't have much success.’
Hugh Foss may have been a brilliant codebreaker but his eccentricities were legion both at work and at home. His cousin Elizabeth Browning, who was now also working in the naval section, was a regular visitor to his home.
I saw a lot of Hugh and Alison Foss. They lived in a bungalow at Aspley Guise with their two small children, one of whom was my god-daughter. The house was always chaotic, as Hugh's wife was a darling but almost totally incompetent domestically. Hugh went home pretty well every day at four-thirty in order to put the children to bed, get supper, and do what he could to organize things. An example of their modus vivendi was the highly complicated arrangement for washing-up (dreamed up, needless to say, by Hugh). Every article was supposed to be washed in a particular order – saucers first (as least polluted by human lips); then teaspoons; then sideplates; then pudding plates; soup bowls; main-course plates; knives; glasses; cups; forks; pudding and soup spoons; and finally saucepans. As these were usually stacked on the floor the dogs were a great help. The theory of this procedure in the days before dishwashers may have been excellent but in practice one usually found two or three days’ washing-up waiting to be done, with plates and dishes piled around and in the sink. If one tried to help there would be shrieks of ‘Oh you mustn't do the cups yet. Saucers first.’ There was also in theory some weird arrangement so that things Hugh was supposed to put away were located at distances appropriate to his great height and long arms, while Alison, who was small and dumpy, had a shorter range. But in practice things ended up pretty well anywhere. I remember having lunch there one day with Hugh's muddy boots on the table beside me to remind him they needed cleaning.
Like most of the civilian codebreakers, Jon Cohen and Hugh Denham were billeted in a private house.
Hugh and I were in a village called New Bradwell, which was attached to the railway company. The house was run by a dear old lady who must have been about seventy and was the widow of somebody who had worked on the railway. It was just a two-up, two-down cottage. I was from an upper-middle-class background in London and this was far in the other direction. But Hugh and I got on very well there. It was wartime and during the war people did all sorts of things they wouldn't always think of doing.
There were also very considerable class differences at Bletchley Park itself. I took up with a girl who I was quite surprised to find was a countess's daughter, because with my middle-class Jewish background that wasn't the kind of person I would normally mix with. But it was a place where all sorts met and there were dances and parties and we enjoyed ourselves to a certain extent. But there was always the background and the need-to-know criteria, that i
s to say you didn't ask questions about what other people were doing or working on. You never went beyond your own narrow field.
Bletchley Park was full of eccentrics and there was a very informal approach to rank and the wearing of uniform, Cohen recalled.
One day the military police guarding the entrance to Bletchley Park saw two RAF sergeants walking down the driveway. They suddenly seemed to stop, look around them and walk very fast in the opposite direction. It was thought that this was suspicious behaviour and they were arrested and brought into the guardhouse. It turned out that they had a quite valid posting to Bletchley but they didn't like the look of it because people were there walking about, officers and ratings, people in uniform and people not in uniform, all sorts, all gesticulating and arguing and so on. So they thought this must be a services lunatic asylum. They said to each other: ‘Well, knowing what postings are like in the armed forces, this is a mistake. We'd better get away quick or we shall be mistaken for a couple of the lunatics.’ It was explained to them of course that it was safe and they came in.
Meanwhile, at Kilindini, poor reception, an acute shortage of personnel and the point-blank refusal of Fabian to exchange results meant that the British codebreakers had made no progress at all with JN25c.
A procedure had been set in place whereby Fabian's unit in Australia, now known as Fleet Radio Unit Melbourne (FRUMEL), and Kilindini would exchange results via Nave in Melbourne. But the veteran Australian codebreaker had fallen out with both Fabian and Newman, who had accused him of breaching security by helping the army codebreakers at Central Bureau. The belligerent Fabian blamed security concerns over Nave for his reluctance to exchange material with Kilindini. Given the American's admitted predilection for only exchanging material where his own unit was likely to gain from it, this may be only part of the truth. Starved of information by their US counterparts, the British codebreakers lost complete touch.
Meanwhile, through a combination of good organization, close co-ordination and extensive use of tabulating machinery, and no doubt boosted by the effect of codebreaking on the Battle of Midway, the US codebreakers forged ahead. The Kilindini code-breakers were devastated by a signal received from Bletchley Park in late July. It was an update from OP-20-G on US progress against JN25c which came only five weeks after Kilindini had estimated it would take six to eight months to begin to read messages in real time and stated that:
Stereotype movement reports of small units and certain type of intelligence reports on our communications in Pacific area are already readable but these types only occasionally yield intelligence of any value. Convoy and shipping control reports should be sufficiently readable within a fortnight to disclose any important activities which may occur in that field. Difficult to estimate probable progress in reading traffic dealing with projected operational large-scale movements, etc. But believe within month general context of 30 per cent of such messages can be determined with perhaps 15 per cent yielding specific information. Forthcoming succeeding months should see 50 per cent increase over preceding months in each of the above categories.
Realizing how far behind they were, some of the British code-breakers, led by ‘Bouncer’ Burnett, suggested they would be serving Somerville better if they moved en masse to Washington. They believed Bletchley Park was only concerned with the European War and as a result was starving them of resources, while the Americans seemed uninterested in intelligence covering the Indian Ocean areas patrolled by the British Eastern Fleet. A British codebreaking presence at OP-20-G would ensure that Somerville received the information he required.
Shaw, who as head of the unit had a better grasp of the logistical problems – not least the major manpower crisis at Bletchley Park – agreed with the Admiralty that the answer lay in a separate, bigger British operation, but with a direct link to Washington. However, Burnett had Somerville's ear, and amid a damaging clash of personalities Shaw found his position undermined.
Somerville appears to have sought a rational compromise, sending Burnett to London in an attempt to sort things out, and in particular to press for the direct radio link between Washington and Kilindini which would have solved most of the problems. Somerville suggested sensibly that a British naval codebreaker should be attached to OP-20-G, adding that ‘Burnett would be eminently suitable in a technical capacity as he has a brilliant brain and is 90 per cent responsible for any results we achieve at Kilindini. But he is somewhat intolerant and might not prove satisfactory if used for purely liaison duties.’
Although Somerville seriously overestimated Burnett's contribution to the British codebreaking operation, the assessment of his protégé's personality traits was prescient. Burnett's part in the subsequent debate were severely limited by an early clash with Commander Edward Travis, the head of Bletchley Park. The British had long pressed for America to take overall responsibility for Japanese codebreaking, leaving them to control the German and Italian operations. But they were not prepared to be entirely dependent on Washington for any area where British servicemen were fighting.
Travis had already ordered a major expansion of Bletchley Park's Japanese codebreaking operations. ‘It became evident that a more vigorous policy was being pursued at GC&CS,’ Shaw said. ‘It displayed more practical interest in Kilindini's work, arranged communications via the RAF in East Africa, supplied Typex machines and offered more personnel.’
At the same time, the British codebreakers enjoyed their first major success since leaving Colombo. The Japanese Merchant Navy used its own special codes or ciphers to communicate with the Imperial Navy and to protect shipping movements. In May 1940 it had introduced a code which the Americans dubbed JN39. Within six months the Allied codebreakers had managed to crack the relatively easy code. But in February 1941 it was changed. ‘The FBI, thinking that they were being helpful, had boarded a Japanese freighter in San Francisco harbour and stolen a copy,’ said Joe Eachus, a former US Navy codebreaker. ‘The first thing the Japanese did when they found out was change the merchant shipping code.’
Its replacement, JN40, was believed to be a code superenciphered with a numerical additive in the same way as JN25. But in September 1942 a textbook error by the Japanese gave John MacInnes and Brian Townend, another of the civilian codebreakers sent to Kilindini by GC&CS, the way in.
The Japanese operator omitted a ship's position from a detailed message and instead of sending it separately in a different message, re-enciphered the original with the same keys, this time including the longitude and latitude that had previously been missing. A comparison of the two messages made it immediately clear that JN40 was not a superenciphered code but a transposition cipher. It was based on a daily changing substitution table, containing 100 two-figure groups or dinomes, each representing a kana syllable, a romaji letter, a figure or a punctuation mark. The operator wrote out the message in kana syllables and then substituted the relevant dinomes. This produced a long sequence of figures which was written into a 10 ¥ 10 square horizontally and then taken out vertically, thereby splitting up the dinomes and making it more difficult to break.
Within weeks MacInnes and Townend had discovered that the substitution tables were formulated in a predictable pattern. By November the codebreakers were able to read all previous traffic and be confident of breaking each message in real time, allowing enemy supplies to be tracked and attacked at will by Allied submarines. What was more, since it was a cipher, there were no code groups to recover and therefore no gaps in any of the messages. ‘This was the first time that any large body of non-coded naval Japanese had become available,’ said MacInnes. Over the next fortnight, they broke two more systems. The first was the previously impenetrable JN167, another merchant-shipping cipher. The second was JN152, a simple transposition and substitution cipher used for broadcasting navigation warnings.
Meanwhile, the first steps towards resolving the problems with JN25 and the difficulty in getting anything out of Fabian were being taken by Edward Travis and Frank Birch, he
ad of the Bletchley Park naval section. During a visit to Washington in September and October 1942, they thrashed out a co-operation agreement on naval interception and codebreaking with OP-20-G. This was dominated by the continuing dispute between the two sides over naval Enigma, in particular the British attempts to break the four-wheel version used by the German Navy's U-boats, dubbed Shark.
Motivated primarily by a determination to protect the Enigma secret, the British were anxious to keep control of the breaking of all Enigma traffic. The US Navy was equally determined to get to grips with the Shark traffic. Although OP-20-G had already begun working on Shark, an agreement on this had been put on hold. It seemed to be the obvious trade-off for full assistance on ‘the Japanese problem’.
But the negotiations did not go as planned, with some very obvious ‘stickiness’ on the American side. Despite being head of the British codebreaking operation, Travis was treated in a very dismissive way and Birch was refused entry to the OP-20-G intelligence-reporting section. The US Navy codebreakers were deeply, and quite wrongly, suspicious that the British were holding back on them. No doubt stung by the British criticism over the lack of security evident in the Midway leak, they had found their own alleged evidence of poor British security in the shape of Fabian's claims against Nave.
The American delegation refused point-blank to allow the British codebreakers at Kilindini to have the secure cipher machines necessary for a direct link with Washington. Under an agreement between the US Navy and the US Army, the American ECM enciphered teleprinter machines could only be used if controlled by US officers. The obvious answer was for a US liaison officer to be posted to Kilindini, but despite the presence at Bletchley Park of a US Navy liaison officer, Washington was adamant not only that none could be sent to Kilindini but that the US Navy representative at Bletchley could not deal with Japanese matters. ‘It always struck me that the main reason why OP-20 would not send us representatives on the Jap side and forbade their European representatives to discuss Jap matters with us was just pride,’ said Birch. ‘On the European side we had led, and they had sent representatives to us, not we to them. Therefore, however different the Japanese problems, it would have been infra dig for them to send representatives to us on the Jap side.’
The Emperor's Codes Page 18