The Secret in Building 26
Page 22
The Japanese attaché also provided helpful information on the new Schnorchel-equipped U-boats, with their ability to “breathe” underwater and stay submerged for longer periods of time. However, none of the attaché’s scientific advisories can compare to what became one of Ultra’s most important contributions to the war: his detailed description of the German fortifications for repelling the D Day invasion.
In 1944, OP20G’s codebreakers also discovered a special Enigma that Japan was using, called the T machine. By midyear, work began on a radically new Bombe, named Bulldozer because of the broad statistical nature of its attack, for use against the Japanese device. Although looking much like the double Bombe, the Bulldozer was of a new breed that did not rely on cribs. Instead, it used advanced electronic circuits to see if its many Enigma banks were producing plain language from the cipher text as it plowed through all the possible wheel positions and combinations. Bulldozer was less elegant and less efficient than the Turing menu approach, but it got the job done—taking hours, not minutes, for a single test. Although not in operation until 1945, Bulldozer’s application of electronics to statistical methods was a breakthrough for the Americans. *31
13
New Challenges . . . and Breakdown
February 19, 1944—Washington, D.C., and Dayton, Ohio
EDWARD TRAVIS’S TELEGRAPHED request to Joseph Redman, director of U.S. naval communications, seemed too much to ask in early 1944 even in the context of the close relationship GCCS and OP20G had built over the past three years. As the head of GCCS, Travis explained that the British had learned of German plans to make the four-wheel Enigma a standard for all naval and U-boat systems. With the upcoming Allied invasion perhaps just months away, Travis continued, Bletchley Park might be overwhelmed by a flood of messages and new keys to be solved each day. The British could become locked out of some of their most valuable long-term sources of information at a critical juncture in the war.
On February 19, Travis asked that OP20G build at least fifty more four-wheel Bombes for use in Washington but in service of British codebreakers. Travis may have been unaware that Desch’s team had switched to other high-priority work and that the Navy might find it hard to get the funds and workforce to produce more new Bombes, especially so near the planned invasion.
Although with misgivings, the Navy’s highest echelons acceded to the British request just two days later. Admiral King ordered OP20G’s research team, NCR’s management, and Desch to start a new multimillion-dollar Bombe project to build, as quickly as possible, fifty updated machines. OP20G’s researchers immediately began exploring the logic and mathematics of possible improvements to Desch’s original design.
Desch must have been beside himself. He already was pressed by labor shortages and the development of a radically new codebreaking machine, Duenna—one of the few projects during the war that would enable Navy researchers to get ahead of the technological curve in cryptanalysis. Work on Duenna began as early as July 1943, when Engstrom told his mathematicians to think of attacks and machine designs for a much more complex Enigma, based on fears the Germans would change to a fully changeable reflector. That, fortunately, did not happen, and Engstrom’s group was shifted to other tasks in 1943. But some important preliminary work on the problem had been done.
The Duenna project was revived in February 1944 when OP20G began to worry that the U-boat Enigma would be refitted by summer with a fully pluggable reflector—that is, a fourth wheel whose internal wiring the operator could change by inserting plugs into different letters. The Allies called the new German device “Uncle Dora,” or simply Uncle D, an innocuous label for a device that would add so many more ciphering possibilities that it would shut out the American Bombes completely.
A fast enough attack against Uncle D demanded a powerful new machine using the latest electronics. By early February 1944, the first design for Duenna was ready. For a backup, Engstrom ordered additions to the old Bombes that might help them decipher Uncle D messages once the reflector wiring had been identified. The problem was so worrisome that GCCS and the Army launched their own projects for attacking it as well.
By early March 1944, Engstrom declared Duenna to be of the highest priority, and yet precious resources were now being diverted to the development of fifty more Bombes. By May, Engstrom and Wenger began to protest loudly against the British request—their engineers were swamped with other projects. They also might have believed that the British were not giving their all to their own four-wheel Bombe effort, a result of Britain’s failure to communicate how much of their limited manufacturing resources were being strained to produce all types of codebreaking machines. What’s more, the British were still worried about the reliability of their own four-wheel Bombes. The development of the new Bombes, they believed, was best left to the Americans. *32
Regardless, Engstrom and Wenger asked that the British request be either terminated or scaled back. The ill feelings engendered on both sides of the Atlantic led to a rehashing of some old accusations between the two allies and a bold attempt to rewrite the history of British-American relations.
IN THE MIDST of this new friction, both the British and the American codebreakers began to think ahead to how their superiors and future generations would view their achievements and their collaboration during the war. The British concluded that OP20G would be unlikely to cooperate in writing an objective history of the Battle of the Atlantic after the war. Their instincts were right: Engstrom and Wenger were soon asked by their superiors to write a quick history of their own on the U.S. Bombe project and British-American codebreaking relations. The Navy’s motives were political—and perhaps part of a larger negotiation between the two nations’ intelligence agencies.
Only Wenger and perhaps Engstrom knew how much information and technical assistance the British had offered the Americans in their 1941–1942 exchanges. Indeed, in a May 13, 1944, memo to his Navy superiors on the history of that period, Wenger had no complaints about the degree of early British cooperation. He stated that GCCS had given OP20G what it had asked for and needed, with the exception of a physical copy of the Enigma, which would not have made a difference anyway. In the same memo, Wenger wrote that Engstrom knew the extent of the British contributions as well.
But less than two weeks later, Wenger, Engstrom, and Meader wrote an internal history of the U.S. Bombe project that blamed the British for the serious delay in OP20G’s ability to understand the workings of the Enigma and create its first Bombe. The history accuses the British of withholding information early in the war and of breaking repeated promises in 1942 to produce a working four-wheel Bombe of their own.
There is no direct evidence to explain OP20G’s sudden about-face, but the historical context points to the need for a scapegoat. OP20G’s research group was in a difficult political position in May 1944 and needed to defend itself to Admiral King and the American naval hierarchy as to why it was resisting the British request for fifty additional Bombes. The internal history also provided OP20G with a one-sided explanation for why the original NCR Bombes did not appear until autumn 1943 and the agency’s anti-Enigma capabilities did not mature until almost the end of that year. The British would have been livid if they had been shown the document.
Tensions between the Allies again ran high, but the success on D Day went a long way toward improving their relations. A few weeks after the Normandy invasion in early June 1944, as Allied bombers and resistance forces cut landlines and forced the Germans to communicate almost exclusively by radio, good cribs for the Army and Air Force codes became plentiful. The two allies realized they would need fewer codebreaking machines and reached a compromise on Britain’s Bombe request. As the latest version of the American Bombes were sent from Dayton to Washington in August 1944, it was decided to produce an additional twenty-five rather than fifty for the British. This still must have created workforce and production-planning headaches for NCR, but not so many that the compromise was unworkable.
Duenna turned out to be a much tougher task for OP20G and Desch—as well as for Bletchley Park and the Army’s codebreakers. Although the Army built two exceptional devices, the Autoscritcher and the Superscritcher (the latter being one of the largest electronic machines of the war), both precomputers came too late to help the war effort. GCCS commissioned two less radical machines, the Giant and the Ogre, both variations on the original Bombe’s architecture and technology, but they proved far less powerful than the Navy’s Duenna.
The completion of Duenna failed to meet the expected August 1944 crisis in the Atlantic by several months. Desch and his team did not deliver the first model until September 1944, and engineers in both Washington, D.C., and Dayton labored feverishly to work the bugs out for a second model. A Duenna machine was not in operation until November.
Despite its late arrival, Duenna was a technological triumph for the Navy: it embodied an innovative and complex version of the no-contradiction tests that were used in the Bombes and could handle menus of eighty letters or more at high speed. Because of the complexity of the tests, Duenna had to be much faster than the Bombes and make more complicated decisions.
Desch and the team in Building 26 had hoped to build Duenna using some three thousand electronic tubes but had to settle for reasons of economy on a combination of traditional relays and three hundred or so electronic tubes, along with a new version of the Bombe’s commutator wheels. The machine worked well, but a single test took hours, not minutes. The time to prepare the special menus for a run was formidable, sometimes taking more than a week.
Fortunately for OP20G, Uncle D never appeared on the U-boat networks, and it was used on only a handful of German Air Force and Army systems. The Germans compounded their mistake by sending the same messages on other systems, allowing easier entry into Uncle D once those messages were broken.
Uncle D’s pluggable reflector was one more item on a long list of technical and procedural improvements to the Enigma that the Germans squandered through poor planning and piecemeal introduction. “It is not easy to see why the device was not used more effectively,” notes American historian Philip Marks.
Despite the difficulties of distributing [Uncle D] while the Germans were retreating on all fronts, they succeeded in deploying large numbers by August 1944 or even earlier, but then failed to use them on a commensurate scale. No doubt passive resistance from operators was a factor, but unclear instructions on when [the device] should be used also played a part, perhaps leaving too much discretion to field units.
Given the German reluctance to use Uncle D, the expense of constructing a Duenna, and the time needed for a run, OP20G decided to build just three of the machines rather than the ten originally planned. Even so, the Navy and Desch’s team could take pride in having been the first with the best.
Designing and building both Duenna and a new Bombe took their toll on Desch in the summer and fall of 1944, and the strain was clearly evident to his colleagues. Desch and his top engineers were now working from 7:30 A.M. to 2:30 A.M. of the following day—a grueling schedule that lasted nearly six weeks, according to Carmelita Bruce, whose husband, Ralph, had been one of the original seven engineers on the NCR research team. “I remember my husband saying at the time, ‘Joe Desch is not doing well. We hope we’ll get him through before all this is over.’ ”
June 4, 1944—South Atlantic
TWO DAYS BEFORE D Day, Admiral Dan Gallery’s hunter-killer group of antisubmarine vessels was about to end its patrol off the coast of Africa when one of its destroyers thought it had finally located the elusive sub that Gallery had been chasing with the aid of Ultra for several frustrating days: U-505.
Launching planes from his escort carrier, the Guadalcanal, Gallery guided his other ships to the location of the submerged U-boat. Depth charges severely damaged the submarine, forcing its wounded skipper, Harold Lange, to surface. Unable to fight off the destroyers, Lange watched as his submarine was boarded by U.S. sailors and its contents seized. Soon, it was towed to an Allied port.
Among the items captured were the keys for the Atlantic and Indian ocean U-boat systems for the first weeks of June. The timing of the seizure was a mixed blessing. With the keys in hand, the Allies might not need to run their Bombes on the two systems for much of June. On the other hand, if Dönitz learned of the capture of U-505, then new keys would be issued and, perhaps, extra Enigma security procedures taken that might lock out OP20G and GCCS during the critical days of the Normandy invasion. The British voiced such concerns to the American Navy.
Luck again was with the Allies, however. The German systems were not changed, and, after the captured settings were delivered to the codebreakers, the Allies found they did not have to use the Bombes for Shark or the Indian Ocean U-boat traffic until the last week of June. Hence, during the critical first weeks of the invasion of France, the Bombes were freed for use against the German Army and Air Force systems.
According to an OP20G report, “Almost the entire capacity of OP20G was devoted to Army and Air Force work [in June 1944]. This resulted in a considerable gain in intelligence during a very critical phase of the invasion of France.” The German order of battle and critical bombing targets were revealed by Ultra throughout the invasion period. With the increase in cribs from the loss of landlines after D Day, “the volume of decodes was so much greater than previous expectations, that the section charged with their dissemination was virtually swamped,” the report concluded.
OP20G gained a greater role in naval Ultra during the invasion period in the summer of 1944 and held on to it through the remainder of the war. A year before, the Navy codebreakers had been working on only two major systems. At the end of 1943, they were participating in breaking a half-dozen naval-related keys. By June 1944, that number had doubled.
OP20G seemed at last to have surmounted Shark as well. With average decrypt times less than sixteen hours by late summer, Wenger and Engstrom thought that they could shift their men to long-term codebreaking solutions. In August 1944, as they laid out new organizational plans for OP20G after the expected victories over Germany and Japan, they asked their young theoreticians to design an all-purpose, electronic “computer” that would advance the Navy’s codebreaking research, if not its operational capabilities, well into the future.
But the respite for OP20G did not last long. In September, it began losing its command over Shark, despite the new Bombes and the increased skills of its codebreakers. The Germans were pulling their wolf packs out of harm’s way and relying on lone hunters until newer technologies could be introduced to help protect them. The resulting lack of messages and cribs more than doubled the time to find Shark keys. The situation, Wenger reported, was bad and would probably get worse. With fewer U-boats at sea, the Germans were cutting back on weather messages and other “short” signals that had been excellent sources of cribs and locations throughout much of the war. In particular, the Bay of Biscay weather messages, which had proved so vital in breaking the German codes just prior to D Day, would now dry up as the Allies overran France and drove the U-boats from their safe harbor in the bay.
Wenger asked Engstrom and his research group to return to operational problems and focus on finally devising a noncrib, statistical attack on Enigma. Meanwhile, OP20G’s operational problems multiplied as the Germans split up more submarine keys and maintained greater radio silence. October’s keys took more than forty
-five hours to find. November and December took almost as long.
In late 1944, OP20G had to focus on yet another Enigma crisis, one that threatened to wipe out the exceptional power of the Bombe’s Diagonal Board. A surprising Nazi innovation to the Enigma, the Uhr box, had been spotted by the British a month after D Day. The Uhr box was an attachment to the Enigma that changed the stecker wiring with just the turn of a knob, allowing a new stecker to appear several times a day. Worse for Allied codebreakers, the new mechanism could make the steckers nonreciprocal, so that A steckered to B would not necessarily mean that B was steckered to A. Eliminating reciprocal steckers would neutralize the effectiveness of the Diagonal Board and reduce the power of both the British and American Bombes by more than half.
But the Germans again made the mistake of introducing the Uhr box in a piecemeal, unplanned fashion—using the attachment on only a few of their networks and transmitting the same messages in many other networks, thus giving the Allies a plentiful supply of cribs for attacking it.
Other Enigma changes, however, proved unconquerable, and only the German reluctance to make a wholesale switch to the new code systems saved Ultra from disaster in the final months of the war. One such challenge, a system known as Sonder, was first noticed by the British in November 1944. It was a limited new submarine network that apparently changed the selection of Enigma wheels and steckers more frequently. The small number of messages on the system compounded the decryption problem.
A February 15, 1945, memo circulated among Navy codebreakers expressed their continuing frustration with the new code. Although the Sonder keys were thought to be of high importance, OP20G had succeeded in breaking only two Sonder messages since its introduction more than three months before, and only because both intercepts had included portions of a message already broken in the Offizier key, the doubly encrypted cipher used exclusively by German naval officers. The memo stated: