by Jim DeBrosse
* 11That original 1940 machine was even less efficient at testing for the setting of the Enigma’s steckers. Soon after, the British upgraded the Bombe by including an automated stecker test when the Bombe stopped. The addition was called a Machine Gun because of its staccato sound as it searched the stop data. The new Standard Bombe, as it came to be called, was put to work on Air Force, not naval, problems because too little was known about U-boat communications to make the Bombes useful. But even the Standard Bombe was slow. The speed of its thirty-some spinning commutators had to be restrained because the electrical relays that sensed a hit were sticky and slow.
* 12Besides the Enigma aids, Weeks also carried away a large packet filled with documents concerning the Russian, Vichy, and Italian codes, German merchant-marine ciphers, bare bones documents on German naval systems, and, perhaps, some old Enigma keys to allow OP20G to practice Enigma decryption.
* 13The relationship was still far from candid. The American visitors remained unaware that the British, so gracious in accepting their gifts, thought little of most of what they had been given, with the exception of Purple. And out of fear of providing the Americans with a bargaining chip, the British did not reveal their admiration for IBM, which they saw as a possible manufacturer of English-designed and English-controlled codebreaking technology.
* 14The British were guilty of their own leaks. During his visit, Currier had been “stopped by a roadblock in a small village, when the local constable saw two men in civilian clothes, obviously not British, riding in a War Department staff car.” The constable “reacted quickly and asked if we would ‘mind getting out and accompanying him to the police station.’ This infuriated our diminutive Scottish driver who jumped out and confronted the policeman: ‘Ye can nae do this, they’re Americans on a secret mission.’ ”
* 15There were several types of catalogs, but all were limited in usefulness because they depended on prior knowledge of the inner workings of the encoding machines they attacked. The most typical catalog was a complete list of a machine’s possible encryptions for a common word such as “Eins” (“one,” in German). An analyst looked through the list to see what settings had produced his “Eins” crib. A more complex and statistical approach was to encrypt a common letter, such as I, at all settings of a known machine. Then an analyst tallied the frequencies of encrypted letters and compared that distribution to those in the catalog.
Driscoll’s anti-Enigma catalog used another variation, the encryption of letter pairs. Unfortunately, catalogs ran to millions of entries. Driscoll’s catalog for finding just the starting positions of the Enigma wheels filled a room twenty-four by thirty feet.
* 16He may even have revealed that England had finally made the breakthroughs that were allowing relatively constant reading of the German naval systems. At the very least, he must have explained to her that Bletchley had explored the same letter-pair catalog approach years before as a full Enigma attack—and failed—before deciding to use it only as an aid to the Bombes and Banburismus.
* 17She did not leave a helpful and detailed record of her special 1941 attack. She kept her crew at work on the catalog even after Turing sent her a long memo in October 1941 full of understated criticism of what he understood her method to be, along with a not-too-guarded demand for the details of her approach she had promised in August.
* 18Denniston and his colleagues declared they had fulfilled all the agreements and were not holding back vital information. They fired back with accusations of their own: Driscoll had not responded as quickly as she had promised; she wasn’t fully sharing her secrets; and the Americans had waited a month before telling them that Denniston’s information on the Enigma had not been received.
On November 27, 1941, British captain Edward Hastings wrote Denniston from Washington, “There is grave unrest and dissatisfaction in free exchange of special intelligence. . . . Noyes is in a mood to withhold further information unless he receives full reciprocal information on European work.”
* 19In the first version of the Bush design, two strips of punched tape were overlapped and run past each other at very high speeds. The two tapes were shifted (“stepped”) one letter at a time until all possible matches of text had been searched. Wherever the text in the two messages lined up, focused light would penetrate the punched holes and their energy would be registered by the photocells. Electronic tubes would then track the number and locations of the light pulses. The logic of the machine was similar to Turing’s Banburismus, but both methods were developed independently.
* 20Three separate attempts were made to design and develop a RAM Comparator for codebreakers, using a variety of punched tapes and microfilm. And even though the electronic tubes for counting matches remained slow, bulky, and temperamental beasts as late as 1940, they turned out to be the least of the problems for the machine’s design. The biggest and most resistant obstacles were mechanical: the overlapping tapes and their punched holes could not be aligned precisely and reliably enough at high speed to permit accurate reading of their matches. The tapes had to be aligned within one thousandth of an inch. Analyzing a thousand-letter message required a million passes by the reading head, through one thousand different steppings of tape, each time with the tapes in perfect alignment.
* 21Having never seen the word “Bombe” in writing or in print—by agreement with the British—Desch might also be forgiven for misspelling the word as “Bomb” in his memo.
* 22The Burger children said they later learned that their maternal grandfather, Friederich Zimmerman, had joined the Nazi Party only to keep his postal job. The children also said that Albert Burger gave his life trying to prove he was a loyal American: he worked extra shifts at General Motors in Dayton during the war, packing graphite powder into bombs, and died from emphysema in 1977 as a result of his exposure.
* 23While IBM could deliver its standard machines quickly, it took them many months to produce the first complex version of the special Navy tabulator equipment.
* 24The British explored a statistical method, based on the patterns and habits of German wheel selection over time, to reduce the number of orders to be tested. They eventually turned away from the method, thinking that the Germans had discovered their own predictability and had taken measures to correct it. But then one of OP20G’s new breed of codebreakers, William Randolph Church, decided to take another look at the German wheel orders. After much labor, he and his crew found a pattern of selection and were able to send, along with the menus from the Crib Group, lists of the most likely wheel selections. This narrowing of the possibilities doubled the efficiency of the American Bombes—saving what amounted to fourteen full months of Bombe time over the duration of the war.
* 25NCR founder John H. Patterson had been a pioneer in employee welfare programs, offering free hot lunches, company-paid trips to the seashore, and picnics on his estate. By the time of World War II, NCR provided employees not only with health insurance and sickness benefits but with medical and dental clinics, a recreation park and pool, and an on-campus office for issuing state fishing and drivers’ licenses. There were free night-school classes, free movies three times a day in the NCR auditorium, and even free umbrellas when it rained.
* 26Naval historians have overestimated the percentage of sub sinkings assisted by Ultra intelligence, claiming it was as high as 70 percent of all U.S. sub kills, in large part because they assumed a direct link between any sunken U-boat and any type of radio or communications intelligence. But a closer examination shows that many sub kills were credited to the aid of codebreaking even when the U-boats were sunk many days or even weeks after the location date given in the U-boat’s message. In other cases, the credit given to decrypts was actually the result of radio direction-finding or, perhaps, traffic analysis. Although the corrected figures are less dramatic, Ultra intelligence still played a significant role in the Battle of the Atlantic.
* 27Some of the limitations of the early American Bombes were spotted soon aft
er their design was set in early 1943. First was the realization that the Bombe needed modifications if it was to be efficiently used as a high-speed version of a “locator”—that is, a means of finding the message setting after other parts of the keys of an Enigma message were known. Borrowing on ideas from the British in the spring of 1943, John Howard outlined the first of the so-called Grenade attachments for the Bombes—a five-foot-long, twenty-pound attachment containing a series of electrical switches. By the end of war, some six different types of Grenade attachments were constructed, each performing a more sophisticated set of functions.
More limitations of the Bombes were recognized in the summer of 1943 when the British warned the Americans that the Germans were about to make changes to Shark. With only a vague idea of what the Germans were up to, Engstrom ordered his mathematicians to think both of new methods of attack and of the outlines for a new type of Bombe. But by the time the first ideas were generated, it was discovered that the changes to Shark were less daunting than thought. The Germans had simply added a new fourth rotor that was to be changed only once a month. Although that prevented the reading of Shark during the first half of July 1943, GCCS soon found ways to identify the new rotors and helped OP20G to master the techniques.
* 28OP20G had been working on Bovril problems by hand and tabulator methods since early 1943 and began running them on the Bombes as early as September 1943.
* 29Much more limited in its application—and developed much earlier than Granddad—was the inverted Bombe. Following British advice, OP20G soon recognized that some German problems demanded that the wheel order in the Bombes be reversed so that the slow rotor became the fastest. That allowed the analysts to deal more easily with rotor turnovers within a crib. By the end of 1943, several of the Bombes had been inverted for such use, receiving the name “Fire Engines.” In addition to standard Bombes, the Fire Engines may have helped with the attack against Bovril.
* 30Breaking into a code system such as JN-25 was both an art and a science, requiring intensive labor. A codebreaker first had to determine the meanings of the code numbers. Unless a codebook was stolen or captured, “book building” was a painstaking process that depended on linguistics, familiarity with a sender’s habits, and much intuition and luck. The next steps were to determine the contents of the book of additives and what number had been added to (or subtracted from) each code group. This method of “differencing” called for hundreds of calculations on columns of code that codebreakers suspected had been enciphered with the same additives. Such an alignment was called a “depth.”
A related method, called stripping, was to try adding likely numbers to the code groups in messages until codes known to occur frequently in Japanese traffic started to emerge. The appearance of high-frequency groups indicated that the codebreaker might have found the right page and starting place of the additives.
* 31Although faced with many demanding assignments during the war, OP20G’s researchers were never asked to deal with the one German system that might have forced them to explore modern digital electronics: the advanced secret teleprinter the Germans were using to encipher their highest-level military communications, a system known as Fish. In a standard Teletype, each letter is represented by a series of five on/off electrical pulses—what computer types call a five-bit binary code, with zero signaling off and the number 1 signaling on. Those simple on/off pulses or “bits,” performed at nearly the speed of light, form the basis of modern computing. A teleprinter might transmit A, for instance, as 11000—meaning two on pulses and three off—and B as 10011. Fish added coding wheels to the teleprinter to automatically encrypt the stream of pulses as they were being transmitted.
If OP20G’s bright young researchers and engineers in Building 26 had focused on Fish and its binary approach to coding, they might have come closer to developing an electronic precomputer. Although there were fears that Fish might be used by the German Navy, dealing with the advanced teletype device was left to the British and the U.S. Army’s codebreakers.
The British concentrated much of their hard-pressed resources for development in 1943 and 1944 on producing machines for the Fish problem. The nature of the attack led the British into designing machines that worked on bits rather than letters. GCCS built two types of advanced machines to attack Fish, named Robinson and Colossus. The Robinsons were high-speed tape machines somewhat like OP20G’s Comparators. The Colossi, first delivered in mid-1944, were more innovative and relied more on electronics—so much so that, although they were special-purpose machines, some view them as the first modern computers.
* 32Between the summer of 1943 and the war’s end, the British increased the number of Bombes in operation from seventy to more than two hundred. They constructed more than sixty four-wheel machines. At the same time, they built ten or more Colossi and a host of other devices for the Fish traffic and for special attacks on difficult Enigma key systems.
Notes
Abbreviations
ACC
accession
ASW
Anti-Submarine Warfare
Crane
Naval Security Group Detachment (Crane, Ind.)
CNO
Chief of Naval Operations
CNSG
Commander Naval Security Group
DOJ
U.S. Department of Justice
ERA
Engineering Research Associates
FBI
Federal Bureau of Investigation
FOIA
Freedom of Information Act (documents obtained via)
GCCS
Government Code and Cypher School
HCC
Historic Cryptologic Collection
HW
records created and inherited by GCHQ (GCCS)
IEEE
Institute of Electrical and Electronics Engineers
MIT
Massachusetts Institute of Technology
NARA
National Archives and Records Administration
NC
Navy Change Machines
NCML
Naval Computing Machine Laboratory
NCR
National Cash Register Company
NCVA
Naval Cryptologic Veterans Association
NDRC
National Defense Research Committee
NOAA
National Oceanic and Atmospheric Administration
NSA
National Security Agency
NSA FOIA
Records obtained via FOIA requests to the NSA
NSGC
Naval Security Group Command
ONI
Office of Naval Intelligence
OP20G
Communications Division of Office of the CNO
OSRD
Office of Scientific Research and Development
PRO
Public Record Office
RAM
Rapid Analytical Machines
RAM File
documents on OP20G’s machine projects provided by the NSA via FOIA requests
RG
Record Group
RG 38
records of the Office of the CNO
RG 457
records of the NSA
RIP
Registered Intelligence Publication
SIS
Signal Intelligence Service, U.S. Army
SRH
Special Research Histories
SRMN
Special Research Materials Navy, Discrete Records of Historical Import: U.S. Navy
SSA
Signal Security Agency, U.S. Army
USDJ
U.S. Department of Justice
USDJ/FBI FOIA
documents provided by DOJ and FBI via FOIA requests
Double quotation marks indicate formal titles for documents. Single quotation marks are for content indicators created by the authors.
Prologue
He was also cold Edward P. Rego inte
rview.
By midnight, a wolf pack Winton, Ultra at Sea, p. 120.
Just before midnight Edward P. Rego interview; Blair, Hitler’s U-boat War: The Hunted, 1942–1945, p. 255; Bowerman and Bridger, “The Battle of the Atlantic.”
In all, fifty-one http://armed-guard.com.
And so it went Morison, Battle of the Atlantic, p. 412.
In ten days Parrish, Ultra Americans, p. 159.
“If a submarine sinks Morison, Battle of the Atlantic, pp. 127–28.
But unknown to the Allies NARA RG 457, Box 112, SRH 368, “Evaluation of the Role of Decryption Intelligence in the Operational Phase of the Battle of the Atlantic”; NARA RG 457, NR1695, as in the Clay Blair Papers, American Heritage Center, University of Wyoming, Box 166, File 11, “German Naval Communications Intelligence.”
Theoretically, at least Miller, Cryptographic Mathematics of Enigma.
With twenty letters Ibid.
Going by the list Sale, “Bigrams, Trigrams and Naval Enigma”; NARA RG 457, HCC, Box 621, ACC7465 CBKJ18, “German Cipher Key Logs.”
The hurdle was too much Rohwer and Beesley et al., “Ultra and the Battle of the Atlantic”; Rohwer, Critical Convoy Battles of March 1943, p. 240; Hodges, Alan Turing, p. 224.
No wonder that Kahn, Seizing the Enigma, p. 217.
To grapple with Shark PRO HW 3/93 ‘Bombe Story,’ and PRO HW 3/93, ‘Wynn-Williams Assigned.’
Progress was slow NARA RG 38, RIP, Box 169, RIP 403, April 21, 1942, ‘Following for Tiltman from Travis’; NARA RG 38, RIP, Box 169, RIP 403, ‘Special British Reports on German Cryptography.’