Codebreakers Victory (20 page)

When, in the autumn of 1929, William became chief of the newly created Signal Intelligence Service, he was empowered to begin building a cryptographic staff. He requested mathematicians, and chance gave him Frank Rowlett, a high school mathematics teacher from Virginia, and two graduate mathematicians from the City College of New York, Solomon Kullback and Abraham Sinkov. Also foreseeing the need for a translator fluent in Japanese, Friedman was swayed by a congressman to hire his nephew, John Hurt.

There were only five of them, but what a team they turned out to be! Friedman adroitly challenged his staff, first by having them study his cryptography writings, then by giving them relatively easy ciphers before moving on to more difficult examples. He threw at them a Great War German cipher that he, with a French cryptanalyst, had had considerable trouble in cracking. The team surprised him by the speed with which they broke it. He turned them loose on the Hebem mechanical enciphering device, which had taken him six weeks of intense concentration to solve; they broke it in less than a month. They were ready, Friedman decided, to take on "real work": breaking the Japanese nonmachine "pencil and paper" codes then current. To Hurt's delight the cryptographers were soon giving him plaintext to translate and interpret. Sinkov recalled of Friedman, "His teaching was such that we developed on our own."

The second heavy responsibility placed on the Friedman five has already been mentioned: the work of developing secure cryptographic systems for the U.S. that resulted in the Sigaba M-134. Because of security restrictions, Friedman couldn't reveal even to his own lawyers the nature of the patents he was seeking. He had to write his own briefs.

 

 

Conquest of the Red Machine

 

As the 1930s unfolded, Friedman and his SIS team faced the challenge of breaking into the code machines the Japanese Foreign Office had introduced. First to be tackled was the Type A machine, the Alphabetical Typewriter 91.

Rowlett, in his memoir,
The Story of Magic,
gave a progressive account of how the Friedman team analyzed the Japanese machine and slowly solved it. They began by observing a quirk in the system. In their new machine ciphers the Japanese used the Roman alphabet to spell out phonetic equivalents of Japanese words, and in this new machine cipher, six letters occurred with high frequency, while the other twenty appeared less frequently. Friedman and his young analysts determined that the six consisted of the five vowels plus the letter
V.
The SIS was able to exploit this discovery by looking for patterns. One such pattern was that in the Japanese phonetics the
Y
was always followed by one of the other vowels and often by a doubling of them—
YUU
or
YOO
—and it was often preceded by
R
or
K—RYUU, RYOO, KYUU, KYOO.
Another pattern was the combination of the letters that, when deciphered, produced
oyobi,
the Japanese word for the English "and." In their analysis the Friedman crew noticed that their descriptions of these identifiable combinations held for only forty-two positions; then the machine introduced a stepping pattern, moving the whole process forward to a new equivalent of the letters. From those beginnings they were able gradually to open up the entire encoding system and produce plaintext from the messages.

Type A, they decided, was a rotor machine, a Japanese variation on the principles of the Enigma. Its encipherment mechanism included two rotors, each of which had twenty-six electrical contacts wired around the circumference of one of its sides. There was also a gear wheel with forty-seven pins projecting from it. A unique feature was that the gear wheel pins were removable. If a pin was in place, it moved the rotors forward with one stroke of the typewriter keys. If it was removed, the machine jumped over that contact, giving the machine an irregular movement meant to foil cryptanalysts.

The machine simply didn't have enough complexity to withstand the SIS team's attack. They were able gradually to open up the entire encoding system and produce plaintext from Type A's messages.

While solving its riddles, the cryptanalysts referred to it often as "the Japanese code machine." They realized that this term was so descriptive that its use might result in an inadvertent security break. They gathered together to settle on a cover name, and eventually the discussion got around to considering colors. "All of us were in agreement," Rowlett wrote, "that the first color of the spectrum was an excellent choice as a cover name for the first cipher machine that we had solved which was actually used for enciphering official messages of a foreign power. And from this moment on, the Japanese diplomatic cipher machine was always referred to as the 'Red Machine.'"

In short order the five team members began to flood the president, the secretary of state and the chiefs of the army and navy with John Hurt's translations of Japanese diplomatic exchanges. Friedman, incidentally, had overcome the usual interservice animosities and developed a friendly, mutually beneficial collaboration with Commander Laurance F. Safford, his counterpart in charge of the navy cryptologists.

Appreciation of the team's exploits took the practical form of increased funding for IBM processing machines, a move to more spacious quarters and an increased staff, including the hiring of bright young women such as Genevieve Grotjan.

 

 

The Breaking of Purple

 

In late 1938 and early 1939, Friedman and his expanded team began deciphering Red messages announcing the distribution of Red's replacement,
angoo-ki taipu B,
or "cipher machine Type B." They hoped that the new machine would be only a modification of Red, but a new series of messages sent in by U.S. intercept stations beginning on March 20, 1939, dashed that hope. They proved unbreakable by the methods applied to Red cryptanalysis.

As the Americans were to learn, slowly and painfully, Japanese cryptographers had built on their experience with Red and had perfected a machine whose principles were completely different from those in the German Enigma or other code machines and presented complexities greater than those of the Enigma.

Again Friedman met with his team to choose a new cover name. Sticking to the light spectrum, they chose the color Purple.

Ronald Clark, in his biography of Friedman,
The Man Who Broke Purple,
told of the felicitous decision made early in 1939 by General Joseph O. Mauborgne, chief of the Signal Corps. Disturbed by the slow progress being made in solving the Purple machine, Mauborgne saw that Friedman was too burdened by administrative details and work on the Sigaba to be effective in the attack. In February 1939, consequently, he ordered Friedman to drop all other duties and concentrate on Purple cryptanalysis. "Friedman," wrote Clark, "now began the most acute eighteen months of intellectual effort he was ever to undertake."

Rowlett's memoir suggested that Friedman had much less to do with the cracking of Purple than did Rowlett and other members of the SIS. However it was, progress did come at a brisker pace. Japanese errors helped. Reliance on ceremonial diplomatic forms of address such as "I have the honor to inform your excellency" handed the analysts cribs of probable plaintext to test against the ciphertext. Purple messages sent to embassies equipped only with the Red machine were obligingly repeated in that code, offering another angle from which to unravel the cipher. And on occasion a sender would admit having erred in one of his settings and re-send the message correctly, supplying the analysts with an insight into the proper setting.

Exacting analysis also showed that Purple offered the same structural opening as Red. Its system included a subsequence of six often-used letters enciphered separately from the other twenty. Unlike in the Red, though, these six could include any letter rather than just the vowels plus
V.
Rowlett drew up plans for a machine that would be an equivalent of the mechanism the Japanese used for enciphering and deciphering the sixes. Friedman, impressed with the plan, recommended that SIS employ Captain Leo Rosen, who had joined the Reserve Officer Training Corps at MIT while studying to become an electrical engineer.

The hiring of Rosen turned out to be another fortunate choice. Very quickly after being briefed on the nature of the Purple machine and of Rowlett's plan for cracking the sixes, he saw what was needed. This was nothing more than the use of standard telephone switchboard stepping switches. The switches were acquired, the "Six Buster" machine was built, and it did its job perfectly.

One of Safford's cooperating navy analysts, young Harry L. Clark, has been credited with raising the question of whether the Japanese, in their new machine, might be using an entirely different enciphering mechanism than those of other code machines. Could it be that the machine was based on the same stepping switches Rosen had applied in the Six Buster? Rowlett, Rosen and the others thought so, but before that question could be answered they had to gain a great deal more knowledge about the groups of twenty letters. Being able to tell where the sixes' letters occurred in a message helped by pinning down likely cribs. What was needed was a much broader base of plaintext equivalents of the twenties. The team made up large worksheets on which they recorded from each day's intercepts the most probable pairings of plaintext and ciphertext equivalents.

The process took most of a painstaking year. Then Genevieve "Gene" Grotjan, who had been doing the monotonous task of compiling indicator worksheets, made the breakthrough. Rowlett has described being in his office conferring with colleagues Bob Ferner and Albert Small when Grotjan broke in, visibly excited. She wanted to show them something in her worksheets. Leading them to her desk, she indicated where she had drawn circles around selected plaintext and ciphertext equivalents. Further, she had circled the same relationships on other worksheets. What she had found were
consistent relationships,
proofs that the coded letters were invariably linked with the plaintexts. As Rowlett characterized her discovery, it was "the first case of positive evidence that we were on the proper course to a full recovery of the Purple machine."

After she had pointed to the last example, Rowlett recalled, "she stepped back from her desk, with her eyes beaming through her rimless glasses, obviously thrilled by her discovery." The others immediately realized the importance of what she had found. "It was a beautiful example," Rowlett wrote, "of what we had hoped our search would uncover."

He described what followed: "Small promptly started dancing around her desk, raising his arms like a victorious prizefighter, and yelled 'Whoopee.' Ferner, who was usually very quiet and not very much inclined to show enthusiasm, clapped his hands, shouting 'Hurrah! Hurrah!' I could not resist jumping up and down and waving my arms above my head and exclaiming 'That's it! That's it! Gene has found what we've been looking for!'"

They made so much noise that Friedman came out of his office and asked, "What's this all about?"

Rowlett led Friedman to Grotjan's desk and tried to let her tell the story of her discovery. But she was so overwhelmed, and by now weeping in her excitement, that he had to take over.

Friedman carefully examined each of the areas and grasped their implications. He saw that with this break his team could go forward with actually building a replica of the Purple machine. "Suddenly he looked tired and placed his hands on the edge of the desk and leaned forward, resting his weight on them," Rowlett recalled. "I pulled a chair forward and offered it to him. Seated, he addressed everyone in the room: 'The recovery of this machine will go down as a milestone in cryptanalytic history. Without a doubt we are now experiencing one of the greatest moments of the Signal Intelligence Service.'"

The team forged ahead toward producing a clone of the Purple machine. Further work along the lines of Grotjan's discovery confirmed what Rosen and his colleagues had suspected: the devices that caused Purple to advance at regular intervals from one encipherment stage to another were stepping switches, both for the sixes' subsequence and for the twenties'. "Rosen," Rowlett wrote, "promptly started to prepare the layout of a complete cipher machine to duplicate the cryptographic functions of the Purple machine."

Postwar comparisons with the real Purple machine verified that Rosen's replica reproduced the original's complex wiring, its use of stepping switches, its intricate plugboard. If ever there was a feat of visionary engineering, this was it: to imagine the inner workings of an intricate secret machine and produce a copy that was so close an approximation of the original that of the several hundred connections in the clone, only two were wired differently from the Japanese machine. As various commentators have noted, the achievement surpassed those of the Poles and the British, who had commercially available models to serve as guides.

On September 25, 1940, just two days after Germany, Italy and Japan had signed the Tripartite Pact pledging all three to mutual support, Rowlett sat at the keyboard of the new machine in Washington's Munitions Building and started typing the ciphertext of a message. "The machine performed beautifully," he wrote in his memoir, "producing a letter-perfect plaintext."

Once again the SIS had opened a secret window through which the U.S. could spy on Japan's diplomats. As we shall see, however, the eventual consequences of the conquest of Purple were to go far beyond listening in to diplomatic chitchat.

Documentary sources make it clear that William Friedman was not the man who broke Purple. He himself said, "Naturally this was a collaborative, cooperative effort on the part of all the people concerned. No one person is responsible for the solution, nor is there any single person to whom the major share of the credit should go."

Nevertheless, the strain of those eighteen months took their toll on him. In December 1940 he collapsed and was rushed to the neuropsychiatric ward of the army's Walter Reed Hospital. He remained there until March 1941 and returned to full-time duty only on April 1. By then additional Purple replicas had been built and distributed to Allied nerve centers. The flow of information meant that both Britain and the U.S. could navigate the twists and turns of Japanese diplomacy.