The next morning, Colonel Sinkov held a meeting to decide what to do with my table. It was logical to put it in our monthly progress report but that had been typed up and given to the courier so it was decided to send it by radio to Washington, a most generous gesture by Colonel Sinkov because we felt it was only the first big step and that in a few days we would disclose a lot more. We also told them about four-weekly change dates.
When there was no immediate reply from Washington, Zach Halpin, the officer in charge of the Central Bureau IBM tabulating machines, had each message of the first period Richard had sorted punched up on IBM cards and they did their own run.
Captain Larry Clark, a veteran of both Rowlett's ‘Purple Section’ and the pre-war Philippines Station 6, examined Richard's findings and pointed out a number of other figures which appeared to relate to each other in some way. If the first and second figures of the fifth group were the same, the first and second figures of the fourth group were the same, too. Clark also noted that the third and fourth figures of these two groups were similarly paired in ‘doublets’. But it did not work the other way round. So if the first and second figures of the fourth group were the same, it was not necessarily the case with the first and second figures of the fifth group. This led Clark to deduce, rightly, that the fifth group was used to encipher the fourth group.
Having made this discovery, Clark went off to lunch, Richard said, leaving the junior codebreaker alone.
I started to copy their list of doublets, hoping to study it later. I soon realized I could not hope to do this in one lunch hour so the thought came to me to try to condense the lists – there were one hundred of them – into one 10 ¥ 10 square table. I remembered from my instructions in Fort Monmouth Cryptanalysis School how Mr Friedman said to recover an alphabet square. To my pleasure, the lists organized into a square which I had out almost completely when Clark returned. He then checked it for accuracy and said, ‘That three-column table of yours should be in this square someplace,’ and after some time and a lot of searching we found the three columns in the square.
By then it was time to leave the office for the evening meal, but after eating I returned determined to recover the column figures across the top of the square. We had the row or key figure but the column or plain-text figures we had not fixed. I was using the three columns from my table as book figures (either 0, 1, 2 or 1, 2, 3) and assuming that the figures of the repeated group with the non-random first figure were respectively the book in use, the two figures of the page number of the additive table and the sum of all three figures by non-carrying addition, since we had seen that order on a captured decoded message in another army system. But the final solution kept slipping away, and I was sleepy. Then Larry Clark called on the telephone and, finding I was there, came in. Together we managed to recover the top coordinates of the square. We finished the task begun in December 1942 on 6 April 1943. What a great feeling to have finally found the solution.
The next morning we sent telegrams to all centres giving the square and all other details. That afternoon we were somewhat deflated to receive a message from Arlington Hall saying they had the square, and proving it by sending the second period square together with instructions for co-ordinating the work on 2468.
Because Sinkov had kept Arlington informed at each step of the way, they had been able to follow in Richard's footsteps, but there was little doubt the credit for the American contribution belonged to him. He could now begin to decipher and decode the messages. With the squares we could recover the additive starting places of all messages and immediately sorted them by book and page, but we found that the row and column co-ordinates on the additive pages were in mixed random order. This meant that in order to write out the messages with the enciphered groups in the correct position above each other, they had to find two or three that started with the same row and column. They slid one message along the top of a second until subtracting a known likely code group in one message gave an additive that, when it was added to the enciphered group below it in the second message, left them with another known and commonly used code group in the second message.
After several possible additives had been found other messages starting in the same row or column could be tried so that other row and column digits could be placed and more messages using them added. Even we non-linguists could start and recover much of a page, starting with the dendai or message number group followed by a number group, which always occurred at least one-third of the way into the message as transmitted. Since the Japanese Army knew that the beginnings of a message were its weakest parts, they cut them into several parts and sent them in a rearranged order. I think this procedure baffled Washington and gave the Japanese Army high-level traffic security for almost a year and a half after the war started.
Apart from the spelling-out of numbers, there was a variety of different cribs that the codebreakers would look for to get into a message, Richard said. ‘There were the code groups for open and close parenthesis, begin repeat and end repeat, begin skip code or end skip code, and most useful of all to the book breakers, begin Chinese Telegraph Code and end Chinese Telegraph Code. I think it was Major Webb who told our linguists to look for that.’
Within weeks, Central Bureau was able to read the Water Transport Code without difficulty and did so until the end of the war. Perhaps more importantly, as a result of the work done by Richard, Noyce and Allen, the Allied codebreakers now knew how the Japanese signallers enciphered the additive indicators for their high-level codes and ciphers. Delhi was already producing decodes which provided details of military units embarking on troop transports for the various areas occupied by the Japanese, and the knowledge gained through breaking it was being used to attack the other high-level military codes.
15
THE YAMAMOTO SHOOTDOWN
The Japanese had made numerous attempts to build up a series of coastal enclaves in northern New Guinea throughout the latter part of 1942. The Japanese Army was forced to pass messages to its commanders on board the transport ships via the easily broken WE code, providing Central Bureau's codebreakers with access to ‘intelligence of considerable operational value’. This and JN25 messages intercepted in Melbourne and Hawaii ensured that many of the convoys were disrupted, but a number did get through. Despite Central Bureau's difficulties in breaking the main army codes, its Australian traffic analysts, using skills honed in the Western Desert, were soon building up a detailed picture of the enemy garrisons at Madang, Wewak, Hollandia and Lae and their efforts to construct new air bases from which to attack MacArthur's US and Australian troops.
In early January 1943 Central Bureau picked up the first signs of a major Japanese attempt to reinforce the garrison at Lae as a prelude to striking south towards Port Moresby in order to capture the tiny Australian garrison at Wau. In response MacArthur and Major-General George C. Kenney, the commander of the 5th Air Force, began a campaign of sustained attacks on the Japanese convoys taking the Japanese 51st Infantry Division to Lae. American B-24 ‘Liberator’ bombers supported by Lockheed P-38 ‘Lightning’ long-range fighters picked off large numbers of supply barges and troop transports, killing thousands of Japanese troops.
As the FRUMEL codebreakers struggled with the Japanese Navy's introduction of an additional codebook, JN25f, in early February, aerial reconnaissance picked up signs of a much larger attempt to reinforce New Guinea. It was not until 19 February that Fabian was able to tell MacArthur that a major reinforcement convoy was scheduled to arrive in Lae in early March. OP-20-G deciphered another message indicating that there would be three reinforcement convoys, the other two going to Wewak and Madang. But the Lae convoy was by far the largest. It was to contain at least six transports and a similar number of escort destroyers, throwing the Japanese 51st Infantry Division into the battle for New Guinea.
In order to disguise the source of their intelligence MacArthur and Kenney sent out reconnaissance aircraft to spot the Japanese convoy. Knowing that it w
ould take five days to reach its destination, they bided their time. The convoy was attacked by a single Liberator bomber on 1 March, losing one troop transport. On the morning of 3 March the US bombers attacked in force in what was to become known as the Battle of the Bismarck Sea. Using a low-level attack technique called skip-bombing, designed to skim their bombs across the water into the Japanese ships, the B-24s sneaked in under the convoy's shield of escort fighters. As the ships caught fire the Japanese troops rushed up on to the deck, only to be caught in a hail of machine-gun fire. All eight Japanese transport vessels and four destroyers were sunk and more than 3,000 soldiers killed. Only 1,000 managed to make it to Lae; the rest were taken back to Rabaul by the remaining destroyers.
The Japanese response, devised by Admiral Yamamoto, was Operation I: concerted aerial bombing of the Allied positions in Papua New Guinea and Guadalcanal. Again, the Allies were informed by Ultra, as high-grade signals intelligence was now known, of what was going on and were able to make defensive preparations.
In mid-April 1943 a number of Allied intercept sites picked up messages referring to a proposed visit by Yamamoto to the Solomon Islands to boost the morale of the Japanese forces based there. It was first picked up in Hawaii, in a JN25 message from Combined Fleet to the outstations that Yamamoto was to visit. FRUPAC's chief linguist, Major Alva ‘Red’ Lasswell, began to work on it. ‘Lasswell approached cryptanalysis like a chess player manoeuvring relentlessly to untangle his problem,’ wrote Jasper Holmes. ‘His desk was usually clear of everything but his current puzzle. He worked sitting upright at his desk, wearing a carefully pressed Marine Corps uniform of the day, his sole deviation being a green eyeshade for protection against the hours under the fluorescent lights.’
As Lasswell worked his way through the Yamamoto message and began to realize what it was about, he called out to his colleagues: ‘We've hit the jackpot.’ He laboured through the night to recover the itinerary. Then Lasswell and Holmes took it by hand to Commander Edwin T. Layton, the Fleet Intelligence Officer.
As the message filtered down to the Japanese lower echelons, it was relayed on in much lower-level systems and was picked up in Australia by both the US codebreakers in Melbourne and 51 Special Wireless Section in Darwin.
Lieutenant-Commander Gill Richardson decoded the message at FRUMEL: ‘The message I worked on was in a Japanese Army code system. Although we were unfamiliar with Japanese Army codes, we got this message out in a hurry because it was a substitution system.’
Nobby Clarke was working at Central Bureau when a message about the visit was picked up by Australian Army operators using a very basic army-air code:
The signal was intercepted by 51 Wireless Section and translated at Central Bureau in Brisbane before being passed on to those thought to need to know. The implications of this message are so bizarre that they have remained clear in my memory to this day. They were sent in the familiar Japanese Army air–ground code they were obliged to use to communicate with their beleaguered outposts. The message was so larded with the kana spellers that anyone familiar with the book could read them on sight.
As the two Betty bombers carrying Yamamoto and his aides flew past the coast of Bougainville, a formation of sixteen Lightnings led by Major John Mitchell of the 339th Squadron USAAF was waiting for them. Four of the pilots were air aces specifically selected for the accuracy of their shooting. While Mitchell and the other aircraft kept the escort of Naval Air Force Zero fighters busy, the four aces attacked and shot down the two Japanese bombers. Yamamoto was found still strapped to his seat wearing a brand-new green tropical uniform and clutching his ceremonial sword. He was cremated and his ashes taken back to Tokyo for a state funeral. The Japanese ordered an investigation amid concerns that the JN25 code had been compromised but eventually concluded that the later use of the low-level code by a junior army commander was to blame. Once again the codebreakers’ secret had survived.
The moves to encircle the Japanese garrison at Rabaul continued throughout the latter half of 1943 with US Marines, under the command of Admiral William F. Halsey but now subordinate to MacArthur's South-west Pacific Area, island-hopping their way through New Georgia, Vella Lavella and on to Bougainville, where they gained a toehold at Empress Augusta Bay. Meanwhile, the Australian and US forces, now under the Australian general Sir Thomas Blamey, took control of eastern New Guinea, capturing the Japanese coastal enclaves at Lae, Salamaua and Finschhafen.
The men of 55 Wireless Section at Port Moresby were monitoring divisional- and regimental-level army networks along the northern New Guinea coast and at Rabaul. But many of these operated on low power, and reception on the other side of the Owen Stanley Range was far from perfect, so a detachment was moved forward to Kaindi, some 7,000 feet above the frontline Australian outpost of Wau, an old gold-mining centre. Based in the deserted home of a former mine manager, the detachment was able to pick up large numbers of new radio networks, many of which used extremely lax security. There were a lot of plain-language transmissions and the detachment was able to report a good deal of intelligence, including details of troops being sent in to reinforce the Japanese garrison at Lae.
Traffic analysis of the operator message logs carried out by Stan Clark's section in Central Bureau built up an extremely accurate picture of the movement of barges down from Palau and along the northern New Guinea coast, allowing General Kenney's 5th Air Force to launch devastating attacks that cut supplies to the increasingly beleaguered Japanese outposts.
When Blamey's forces captured Lae on 16 September and the Japanese 51st Infantry Division began its retreat to Kiari, the Australian Army intercept operators at Kaindi were able to track its movements. Ripped apart during the Battle of the Bismarck Sea as a result of intelligence provided by Central Bureau and FRUMEL, the unlucky 51st now found itself under sustained attack yet again as a result of the codebreakers’ efforts.
Meanwhile, the complete stranglehold enjoyed over enemy air activity by the RAAF operators at Port Moresby had played a major part in assisting General Kenney's aircraft to destroy the Japanese carrier air groups based in Rabaul. The Japanese had lost nearly 3,000 pilots in the struggle for the Solomon Islands alone. Their replacements were not so well trained, aircraft and spare parts were in short supply, and casualties mounted at horrific rates. Japanese naval air power in the region would never be the same again.
The breaking of the Army Water Transport Code led to renewed attacks on other mainline systems. The US Army codebreakers at Arlington Hall were concentrating on the main ground forces’ codes, among them the General Army Administrative Code, or rikugun angoo-sho 4, known as the 7890.
David Mead, a young army corporal at Arlington Hall, recalled how the first breakthrough came in April 1943:
I had asked the IBM unit on numerous occasions for printouts of messages sorted on specific code groups in selected positions in the text. One of these revealed a pair of messages sent on the same date from Tokyo and apparently beginning with the same additive numbers. This could happen by chance, but more likely a Japanese operator carelessly or inadvertently sent almost identical messages without shifting to a different starting point in the additive table. The number of hits or matching groups was much greater than random. Further IBM runs on these matching groups uncovered two more messages, neither sent from Tokyo, using the same table, but with different starting points.
Mead used the depth of messages on the same table to break his way into the 7890 indicator system. ‘As I walked the two miles home that day, I tried to sort out my feelings about the breakthrough,’ he said. ‘Relief, quiet elation, even a tinge of disbelief all combined in a kind of emotional tangle. I wondered, of course, what influence the breakthrough would have on the solution of the code. Most of all, I felt a bond, a sense of community, with the Allied forces who were enduring the hazards and hardships of the war in the Pacific.’
Until now, Bletchley Park's attacks on Japanese military codes and ciphers had been in the hands of
a small Japanese codebreaking section assisted by John Tiltman as and when his other responsibilities allowed. With manpower at a premium, Japanese military had been very much the poor relation of its European counterparts. But with the breaking by the Wireless Experimental Centre and Central Bureau of the Water Transport Code and Arlington Hall's promising assault on the Army Administrative Code in progress, Tiltman decided the time was right for a more sustained and better co-ordinated attack on the Japanese Army high-level codes.
The British called a conference of senior codebreakers from all the main organizations working on the Japanese Army high-level systems. Mic Sandford happened to be visiting Bletchley Park at the time, and he and Tiltman were joined by representatives from Arlington Hall and Delhi. It was agreed that Bletchley Park and the Wireless Experimental Centre should concentrate their cryptographic resources on the codes and ciphers of the Japanese Army Air Force. Arlington Hall would deal with the high-level systems used by the Japanese ground forces, leaving Central Bureau to concentrate on the low-level material produced by its forward field units and the Water Transport Code broken by Joe Richard.
The Bletchley Park Japanese military section was forced to expand rapidly to take on the new responsibility and to cope with the increased intelligence expected as a result of the recent successes. Italy was on the verge of defeat and codebreakers were becoming available for other tasks. The section acquired its own army air subsection in May 1943, and some weeks later a military intelligence section, which reported to MI2, the War Office intelligence section covering the Far East.
The Emperor's Codes Page 22
