by Kahn, David
The orders for the destruction of cryptomaterial were clear and simple; the crews had been taught them, and in all previous cases the men had fulfilled their duty to guarantee security. The fundamental documents were in water-soluble ink and were kept separated: the indicators with the radiomen in the radio shack, the key lists with the radio officer elsewhere. A capture of one or the other would not suffice for solution; both would have to be seized, and this would be possible only if a British warship had, unnoticed, come alongside the German ship and a boarding party had surreptitiously entered her. Aside from the basic improbability of this scenario, Fricke maintained, his survey of the circumstances of the sinking of each German vessel showed that it seems not to have happened. He decided, without stating his ground, that “seizure of cryptomaterial is unlikely.” He made the same determination in every case, either because—as with the tanker Belchen—the ship had been sunk by gunfire and the papers and machine went down with her, or because—as with the Gonzenheim—the crew had had enough time to destroy the material, or because—as with the Esso Hamburg—surprise entry was excluded.
What about pure cryptanalysis, unaided by booty? “After a renewed, very comprehensive examination,” Fricke reported, “all specialists unanimously agreed that a reading [of German navy messages by the enemy through solution] is impossible.” In the end, Fricke found “no palpable, unequivocal” cause of the roundup, though he tended to favor coincidence as the answer.
Despite his failure to reach a definite conclusion, the navy instituted a number of measures to restore or improve security. All orders were to be printed and charts marked in water-soluble ink. All radio messages pertaining to operations were to be top secret. Because documents recovered from the sea might have enabled the British to read German cryptograms for the period that the same keys remained in service, a new cue word—PERSEUS—put new and uncompromised keys into effect on June 22. In August Dönitz began addressing his submarines by the names of their captains instead of by their boat numbers.
The method of defining meeting points by latitude and longitude in the new Short Signal Book was regarded as compromised, so Dönitz sought to disguise their positions on the Kriegsmarine’s grid of the oceans.
This grid was divided into quadrants 486 nautical miles on a side. Each was designated by a two-letter group; quadrants were designated from west to east in alphabetical order. Thus CA covered the East Coast of the United States from about Portsmouth, New Hampshire, south to Cape Fear, North Carolina, and from inland (though this was useless) to about the longitude of northeasternmost Maine; CB, adjoining it to the east, covered the Atlantic south of Nova Scotia; CC, CD, CE, and CF moved east across the ocean, putting CG on the coasts of Portugal, Spain, and North Africa, including the Strait of Gibraltar. South of the C row came the D row. Some quadrants were slightly irregular.
Each quadrant was divided into a nine-by-nine matrix of eighty-one smaller squares. Each of these was in turn divided into nine squares, and finally these again into nine. The squares of the first subdivision, represented by two digits, extended 54 nautical miles on a side; those of the fourth and final subdivision, represented by four digits, were only 6 miles on a side. A vessel could thus give its position with precision using two letters and four digits. For example AK 2799 marked the watery square at 57° 21′ north latitude, 32° 00′ west longitude, a spot in the middle of the North Atlantic where a U-boat might well be in wait for a convoy. The system saved enciphering and transmitting time.
Since the grid served the entire German navy, and since some high headquarters held the Home Waters U-boat Enigma key, these posts could follow the movements of the submarines. Dönitz feared that this situation endangered security. So he took yet another step to protect his U-boats: in addition to forbidding all but a few units to map U-boat locations, he ordered that these locations be disguised by replacing the grid digraphs for the North Atlantic with substitutes, known only to the U-boats and their commands, from Table B of the digraph substitution booklet FLUSS (“river”), also used to encipher indicator groups for the naval Enigma. The table consisted of a 26-by-26 square of letter pairs with single letters at the head of each column and row. This was modified for the grid encipherment. Atop each of the 26 columns the cipher clerk wrote one of the 26 most-used grid digraphs (omitting those for the Pacific, for example) in a sequence specified by U-Boat Command. The encipherer replaced the grid digraph with any one of the 26 digraphs under it. Thus grid digraph AL might become cipher KS, or LK, or OM, or any one of 23 other digraphs.
Dönitz enciphered these instructions in officer-grade Enigma and radioed them in six parts totaling 504 four-letter groups on September 10, 1941, to all U-boats. They went into effect immediately. And at the end of November the navy complicated the location system still more. Dönitz instituted the use of not one but many digraph tables, indicating each by a name and a street address, such as “Gottfried Becker, Bluecherplatz 30,” which came into force at midnight, December 28. The number in the address was the key to the disguise of the four-digit location number. The 30, for example, meant that 3000 was to be added to the true number. Thus the grid digits 6268 would be enciphered as 9268 for transmission. Errors seemed to be rare, though one submarine was told with some asperity that according to its grid letters it was transmitting from the middle of the Andes.
In between these changes, on October 5, the high command segregated U-Boat Command communications from other users of the Home Waters key by modifying the settings. This new key net was called TRITON, after a sea demigod famed for sounding a conch shell as a horn. Not only would fewer people be able to read U-boat messages, but fewer messages would be sent in the basic daily key. This would reduce the chances of overlaps leading to superimposition solutions and of errors that would permit special-case solutions.
Four months after Fricke’s investigation, however, a series of events led to another probe. One was an Admiralty announcement that the British had captured a U-boat all but undamaged at the end of August 1941. This was the U-570, which had unluckily surfaced south of Iceland at the very moment that a British patrol plane was passing overhead. The aircraft dropped four depth charges, which straddled the submarine, shook her severely, smashed many of her instruments, let in sea water, and persuaded the captain that the fight was over. When his men tried to climb out of the conning tower, the plane, to prevent them from manning guns, opened fire. The crew showed a white flag, and the airplane circled watchfully. Eventually a British destroyer arrived and took the U-boat in tow.
On October 18, Maertens opened his analysis of the security consequences of the capture by saying that “a current reading of our messages is not possible.” On the next page, however, he conceded that if the enemy had found the Enigma undisturbed and all the key documents, a current reading was possible. But then he concluded that this was unlikely—that there was time to drench the documents, making their water-soluble ink unreadable—and in the end he left the impression that the British were not solving Enigma messages. Even if they were, the new keys that were to go into effect on November 1 would restore full security.
At about this time Dönitz sensed a trend that affected his tactics in the most fundamental way. He had come to suspect that his wolf-packs were finding convoys less often than individual U-boats were. The inference was that the British were steering the convoys away from the packs. Moreover, the Germans were finding it more difficult to locate the British supply ships. None of this, Stummel implied in a response to Dönitz, meant that the British were reading weak German ciphers. It could all be explained by Britain’s full exploitation of her vast reconnaissance capabilities, including air reconnaissance of the supply route and of U-boat departure and arrival routes; possibly by airborne radar spotting that U-boats could not detect; and by extraordinary British direction-finding. In addition, a spy, a chatterbox, or laxness in locking up maps or cryptomaterial or other documents could not be ruled out. What could be ruled out was
; a current reading of our messages by the enemy. … Without any contradiction, all specialists, in particular those of the B-Dienst headquarters and of the most important specialists of the High Command of the Armed Forces, have determined, in comprehensive work, that the Enigma system is viewed as by far the most resistant of all known methods for secrecy in military communications.
Stummel may have consulted Captain Henno Lucan, the signals officer who in 1930 had proposed some useful improvements to the Enigma. The specialists in the High Command of the Armed Forces worked in the Chiffrierabteilung, or Cipher Branch, the descendant of the ChiStelle whose staff had, in the 1920s, adopted the Enigma for army use. The Cipher Branch’s Desk IVa tested German cryptosystems. Dr. Karl Stein, the desk’s head, a professor of mathematics, analyzed the Enigma theoretically, calculating limits of security and thereby complementing the pragmatic investigations of the B-Dienst cryptanalysts. Maertens summarized Stummel’s report by saying, in his covering letter to Dönitz, that “despite great stresses, including, among others, losses, the resistance of the most important cryptosystems seems not to have been impaired.”
The German cryptologists were not fools. Experience had taught them that cipher systems were usually broken because of laziness or errors on the part of cipher clerks or the capture of documents in wartime; they had not forgotten the Magdeburg. The system they had designed blocked both these avenues and at the same time rendered pure cryptanalysis all but impossible, they believed.
To prevent cipher clerks from choosing rotor settings that might lead to overlaps in the machine’s cipher-alphabet sequence, which might permit solution, they prescribed the settings for a key net. And to nullify capture, they set up the system so that in most cases the capture of even three of the four elements—the machine, the machine-settings list, the indicators list, and the bigram tables—would still (they thought) preclude solution. The machine was assumed to be in enemy hands. But even if the British captured the machine-settings list and the indicators list, the cryptologists said, they would not be able to divine the indicator for a particular message because they did not have the bigram table that would link the unenciphered indicator to the enciphered one that was transmitted. Likewise, if the British had the indicators list and the bigram table, they would not know which rotors had been used in which order, nor their starting positions, nor the plugboard arrangement. Only if they seized the machine-settings list and the bigram table could they reconstruct the indicator and thus be able to read a message. But for this eventuality the Germans had devised the cue-word system. By immediately changing the rotor order and settings, the cue word rendered the captured machine-settings list useless. In addition, the Germans changed the bigram tables from time to time, and new machine-setting lists were issued each month. All these safeguards pretty much eliminated any dangers from captures and from cipherers’ errors, they thought.
The cryptologists believed that trying to solve cryptograms on the basis of letter frequency was laughable: the Enigma generated far too many alphabets, and messages were kept too short, for this procedure to have any hope of working. The more promising method of the probable word, in which a presumed text was matched against a cryptogram, would founder on the vast number of possibilities that had to be tested, and the plugboard would make such a match even more difficult. Moreover, many messages had codewords from the Short Signal Book as their plaintext. And while solving one message would reveal the rotor order of other messages, it would not disclose rotor starting positions.
All in all, the German cryptologists had looked at the situation from both the theoretical and the practical points of view. They had evolved a system that apparently assured nearly perfect security for their messages, in which even a capture would give the enemy insight into German messages only for the brief and limited period before either new keys came into service or a cue word in effect created new keys immediately. Dönitz’s directives to his U-boats, they assured him, were safe.
Nevertheless, the cryptologists did not rest on their laurels. They sought to further secure the system by facilitating the work of the encipherers and thus reducing human error. One measure sought to reduce the pressure—around 5 pounds—needed to depress each of the typewriter keys. The basic resistance came from the rubbing of the rotors against one another. Though this could not be cut down because it would have entailed too extensive changes in the machine’s construction, other changes did lessen the pressure to about 4 pounds, which led to “a palpable lightening of operation.” In another change, a larger illuminable panel was attached to make the letters more easily visible, particularly by the man who was writing them down. A third measure was to print the output, thereby eliminating this second man and his errors. The navy’s first effort, a two-typewriter device called the MS, which weighed more than 100 pounds and cost 5,000 reichsmarks ($12,000 in 1991 dollars), failed. The manufacturers then sought to connect the Enigma to electric typewriters or punched telewriter tape in a succession of “partial solutions” called the MZSB, MZSE, and MZSS devices. But metal shortages as the war went on had led the manufacturers to use plastic instead of metal in some parts, notably the indented thumbwheels for the hand turning of the rotors. This substitution required an increase in tolerances. And despite the Enigma’s heavy construction and simple mechanics, the machine turned out to be extraordinarily sensitive to inaccuracies in manufacture. With an accidental accumulation of variations, the machine with a typewriter-printing attachment became unreliable. As a consequence, though some 700 machines were fitted with the tape printer, none of these “partial solutions” was deemed satisfactory, and the manufacturers, which now included a firm called Konski & Krüger and the Olympia typewriter company, took over the development. Delivery of their printing versions was to start in the fall of 1944.
In addition, the Kriegsmarine sought to reduce the number of machine breakdowns and to meet expanded communications demands by producing more Enigmas. With the original firm, Heimsoeth & Rinke, apparently producing at its maximum, the navy contracted with Olympia. On June 23, 1943, from its factory in Erfurt, southwest of Berlin, the typewriter firm delivered its first twenty basic Enigma machines. By December, it was delivering seventy-five a month.
Most important, the cryptologists also took steps to block the dangers from the constant increase in traffic. In 1939 radio messages averaged 192 a day; by 1942 volume had soared to six times that number, or 1,200 a day; in 1943 it was on its way to doubling again. The experts recognized that this volume gave enemy codebreakers more opportunities. As they put it in one case, “according to cryptanalytic knowledge the permitted limit of the daily total of radio messages had been overstepped.”
One way to lessen the danger was to reduce the number of messages enciphered with the same rotor and ring positions by creating additional key nets. (The participants in each net shared the machine-setting list that specified these positions for each day.) TRITON, the Atlantic U-boat net, was one of these new key nets. Others were created throughout 1942, such as NEPTUN, for the operations of the main fleet, and MEDUSA, for the Mediterranean. By January 1, 1943, the Kriegsmarine was utilizing eleven key nets. As traffic grew, it kept adding others.
Another way to reduce the dangers of high traffic volume was to add another rotor to the machine. The advantages of a fourth rotor, which had been in development since 1940, were that it would raise the number of possibilities a cryptanalyst would have to test and would lower the likelihood of key overlaps that a cryptanalyst could exploit. But two serious practical difficulties blocked the implementation of the idea. First, adding another rotor would change the dimensions of the machine, requiring redesign of and retooling for parts of the machine not otherwise involved and making it impossible, on many ships, for the machine to fit into the space designed for the smaller version. Second, a fourth rotor would prevent communication with other branches of the service that used the three-rotor machine.
To get out of these difficulties, the navy came up
with the idea of a thin rotor that could be fitted in next to a new, thinner reflector (the nonrevolving half-rotor that sent the current back through the three revolving rotors). The thin rotor would not revolve during encipherment because no stepping mechanism existed at the leftmost rotor. However, to create a key, the stationary thin rotor could be turned to any one of twenty-six positions. This multiplied by twenty-six the number of possible keys. And the fourth rotor could be wired so that in a certain position, it, with the new thin reflector, replicated the wiring of the old thick reflector, permitting communication with three-rotor machines. It took more than a year to resolve the problems of the extra rotor and to produce, test, and distribute the new machines. Finally, however, on February 1, 1942, the new model, called M4, went into service on the U-boats’ TRITON key net. This was the most significant event in German cryptography during World War II.
If the advent of the M4 had been followed by a decrease in diversions of Allied convoys and an increase in sinkings in the mid-Atlantic, Dönitz might have guessed that the Enigma had earlier been penetrated. But the decline in Britain’s cryptanalytic fortunes was concealed from him in part by American stupidities. The entry of the United States into the war against Germany on December 11, 1941, voided Hitler’s concerns about sinking American vessels, and Dönitz sent his U-boats to the rich hunting grounds off the East Coast. Here freighters and tankers, trawlers and barges, marched individually up and down the coast, disdaining the lessons of convoy so painfully learned by the British over two world wars. And they did so before a blaze of city lights, foolishly kept burning by chambers of commerce afraid of losing business during the tourist season. As a consequence, for six months the U-boats enjoyed what they called a “happy time,” sinking dozens of ships, sometimes within sight of crowds on shore, with barely a loss of their own. Finally, reason took over, convoying was introduced, and the U-boats quit the coast.