THE CODEBREAKERS

by DAVID KAHN

  He returned from China in 1940, and, after a brief attempt to be a restaurateur in Washington, went to Canada to set up a cryptanalytic bureau which dealt largely with spy ciphers. He was reportedly forced out under pressure either from Stimson, then Secretary of War, or from the British, though the Canadians did not want to part with him. From 1941 to the end of the war he served as an enforcement officer in the food division of the Office of Price Administration. His popular The Education of a Poker Player, in which he offered an informal course of instruction in the game, appeared in 1957. On August 7, 1958, he died of a stroke at his home in Silver Spring, Maryland, and was buried with military honors in Arlington National Cemetery.

  The obituaries called him “the father of American cryptography.” They were wrong, but they demonstrated the deep impression that Yardley’s writing had made on the American consciousness. With all its faults and falsehoods, his book had captured the imagination of the public and inspired untold numbers of amateurs to become interested in cryptology. To the extent that the impact of their fresh ideas enriched American cryptology, the credit must go to him.

  While Herbert Yardley may be the best known cryptologist, uncontestably the greatest is William Frederick Friedman. Unlike his contemporary, his eminence is due most emphatically to what he did. Indeed, two more dissimilar men in a single field can scarcely be imagined. Where Yardley was Rabelaisian, outgoing, superficial, free and easy with the details of a good story, and ever ready for the main chance, Friedman tended toward introversion, depth of study, personal security, timidity, dedication, and accuracy, nicety, and validity of work. Despite the relative drabness of these personal traits—or perhaps because of them, Friedman’s theoretical contributions and his practical attainments exceed those of any other cryptologist. Yardley’s career was like an amazing skyrocket that explodes in fantastic patterns against the heavens. Friedman’s was like the sun.

  He was born Wolfe Friedman on September 24, 1891, in Kishinev, Russia, the oldest son and second child of Frederick and Rosa Friedman. His father, a Rumanian who spoke eight languages and worked as an interpreter for the Russian Post Office, emigrated to America in 1892, at which time his son’s name was changed to William. The family settled in Pittsburgh, where his father managed a sewing machine agency. William graduated in 1909 as one of the ten honor students in a class of 300 at Pittsburgh Central High School; he then went to work as chief clerk in the Erie City Iron Works, a firm that sold steam engines. About that time the back-to-the-farm movement called to city boys, and in the fall of 1910, Friedman and three friends enrolled in Michigan Agricultural College, whose chief attraction was that it was tuition-free.

  But Friedman soon discovered that farming held little interest for him. He was an inventive young fellow who liked to fix things and had written some science fiction for his high-school paper; he was rapidly coming to the conclusion that he liked science. At the end of the term he learned that tuition was also free in a scientific field allied to agriculture—genetics—at one of the Ivy League universities, Cornell. He borrowed train fare and arrived in Ithaca, New York, in February, 1911, where he got a job waiting on tables. After commencement in February of 1914, he attended graduate school, managing to fall in love twice, once with a brunette, once with the blonde daughter of a movie-house owner. While he was there, a wealthy textile merchant, George Fabyan, who maintained laboratories in acoustics, chemistry, genetics, and cryptology (to try to prove that Bacon wrote Shakespeare’s plays) on his 500-acre estate, Riverbank, at Geneva, Illinois, decided that he needed a geneticist to improve the grains and livestock on his farm. He applied to Cornell for a “would-be-er,” not an “as-is-er,” and hired Friedman, to begin June 1, 1915.

  Fabyan was a man of no formal education but of intelligence and energy. He had a great desire to be “somebody,” and that desire motivated his subsidizing the Baconian studies: proof of this revolutionary thesis would cover its patron as well as its actual discoverers with glory. He himself read little, but he absorbed enough from those around him to make his talk on almost any subject sound impressive—at least superficially. He was autocratic, never allowing his staff to disagree with him, but otherwise not unpleasant so long as employees recognized that he was boss. A cardinal article of faith with him was that a well-executed sales campaign could put across almost anything.

  Friedman did some genetics work for him, but, because he was handy with a camera, he helped the cryptologists who were looking for Bacon’s cipher-signatures in Shakespeare by making photographic enlargements of the Elizabethan printing that figured in the work. The Department of Ciphers of the Riverbank Laboratories consisted of 14 or 15 high-school and college graduates who assigned the individual letters in these Elizabethan texts to one or the other of two fonts of type as part of the Baconian search. Fabyan gave them their living plus a salary of about $50 a month. The staff was fed and housed in Engledew and Hoover Cottages, the cipher laboratories taking up the first floor of Engledew.

  The young woman who collated the work of many of the other staff members was Elizebeth Smith. She had been born August 26, 1892, in Huntington, Indiana, the youngest of the nine children of John M. Smith, a dairyman, banker, and county Republican committeeman, and his wife, Sopha, who spelled her daughter’s Christian name with an e instead of an a in the middle because she was not going to have anyone calling her child “Eliza.” After completing high school in Huntingdon, Elizebeth attended Wooster College briefly but was graduated from Hillsdale College in Michigan where she had majored in English. While working at the Newberry Library in Chicago, she was recruited by Fabyan and began work there in 1916.

  Neither she nor Friedman had given any particular previous thought to cryptology, but they began to get personally interested in the work. It is yet another of the ironies of cryptologic history that the interest of two foremost cryptologists was aroused by a false doctrine—a doctrine, moreover, against which they later were to wage a lifetime battle. For at table at the Riverbank cottages they heard gaudy tales of lusty Elizabethan life, of the not-so-Virgin Queen, of courtiers’ intrigues and the secret histories of the great names of English history—all actually invalid decipherments of Shakespeare’s plays tending to prove that Bacon had written them, related by the gentle, upright, but self-deluded woman who had “deciphered” them, Mrs. Elizabeth Wells Gallup. These stories stirred Friedman’s dormant interest; he began to do some of the cryptology, and inevitably its puissant magic seeped like the fume of poppies into his mind and spirit and intoxicated him. “When it came to the cryptology,” he recalled years later, “something in me found an outlet.”

  An understatement. He soon found himself head of the Department of Ciphers as well as the Department of Genetics at Riverbank. The attraction he felt for cryptology was reinforced by the attraction he felt for a cryptologist: the quick-witted and sprightly Miss Smith. In May of 1917 they were married and started the most famous husband-and-wife team in the history of cryptology.

  America had declared war a month before, and Riverbank, which had the only going cryptologic concern in the country, began getting, on an informal basis, cryptograms for solution from various government bureaus. Probably the most important were messages to and from a ring of 125 Hindus who, with German aid, were taking advantage of England’s preoccupation in Europe to strike for Indian independence. The intercepts were given to Friedman for solution, and he quickly solved the number cipher used in cablegrams to Berlin. The letters of the plaintext and of the keyword were transformed into digits by means of a 4 × 7 checkerboard with a normal alphabet; the key digits were then added to those of the plaintext to form the ciphertext. One key was LAMP. Each agent had his own key, but Friedman had no trouble in solving them. Nor was he stumped by a system usually regarded by amateurs as the ne plus ultra of cryptographic security: a book cipher.

  It came to him in the form of a seven-page typewritten letter. The writer, Heramba Lal Gupta, had enciphered only the important words, lea
ving large patches of cleartext as valuable clues; he had also repeated the equivalents for many letters instead of seeking new ones and had employed neighboring letters in a single line, thus enabling Friedman to reconstruct the words of the keytext as a check upon and aid to the solution. For example, Friedman guessed from context that 83-1-2 83-1-11 83-1-25 83-1-1 83-1-8 83-1-13 83-1-18 83-1-3 83-1-1 83-1-6 83-1-3 83-1-6 meant revolution in, with the 83 the page, the 1 the line on that page, and the third number the letter in that line. (It is interesting to note how the third group sticks out as the equivalent for a low-frequency letter by being so far back in the line.) This gave him ORI . . N . L . . E . U . . as the start of the key line, and this in turn probably let him guess that the line started with original or originally. He would then have known that 83-1-4 in the very next word was the equivalent for g in Bengal. By taking full advantage of such clues he built up the entire plaintext without ever knowing what was later discovered—that the key book was Price Collier’s Germany and the Germans, a scholarly work published in New York in 1913.

  The Hindus were prosecuted for trying to purchase the uprising’s arms in the United States and to ship them from the West Coast. At the mass trials in Chicago and San Francisco, Friedman gave evidence that in effect convicted the conspirators out of their own mouths. The San Francisco proceeding witnessed one of the most dramatic scenes ever to occur in an American courtroom when one defendant rose, fired two shots from a revolver to assassinate a compatriot who was testifying for the government, and was himself killed by a marshal shooting over the heads of the crowd. In an anticlimax, a jury later found most of the defendants guilty.

  A few months after these Hindu solutions, the British submitted five short messages to Riverbank for tests. They had been enciphered by a cipher device invented by J. St. Vincent Pletts of M.I. 1(b), the British War Office cryptanalytic bureau. The machine was a modification of the Wheatstone apparatus, proposed as a field cipher. So highly did the British regard it that one argument advanced against its adoption was that if the Germans captured one and adopted it, the Allies would no longer be able to solve enemy messages! Friedman, however, at once recovered the keyword CIPHER to one of the mixed alphabets. But he could not seem to get anywhere with the other keyword and, stymied, he resorted to a bit of psychological cryptanalysis. He turned to the new Mrs. Friedman, and asked her to make her mind a blank.

  How the Hindus worked the book cipher that William Friedman solved

  “Now,” he went on, “I want you to tell me the first word that comes into your mind when I say a word.” He paused. “Cipher,” he said.

  “Machine,” she replied.

  It turned out to be the very key desired. Three hours after Friedman received the cryptograms, their plaintexts were being cabled to London. (The first one read, in a phrase dear to proud inventors, This cipher is absolutely undecipherable.) Needless to say, it ended consideration of the Pletts device for Allied use.

  In addition to this cryptanalytical work, Friedman did most of the teaching of a class of Army officers sent in the fall of 1917 to Riverbank’s Department of Ciphers to learn cryptology. For instruction in these courses, he turned out a series of technical monographs. He completed seven before he went overseas to G.2 A.6 in the spring of 1918 and wrote an eighth on his return. Known collectively as the Riverbank Publications, they rise up like a landmark in the history of cryptology. Nearly all of them broke new ground, and mastery of the information they first set forth is still regarded as the prerequisite for a higher cryptologic education. Fabyan sought to win an implied credit for them by keeping Friedman’s name off the title pages and by copyrighting them in his own. A full set of the white, paperbound pamphlets has become an essential for a good collection of cryptologia, but since only 400 copies were printed, they are extremely rare, and copies of each pamphlet fetch up to $25 apiece on the rare-book market, where they are immediately snapped up on the few occasions that they appear. One zealous amateur thought so highly of them that he painstakingly copied them on his typewriter, and photostatic copies have been purchased by collectors who despair of ever getting the originals. Because Riverbank had issued other publications, the cryptologic series began with No. 15.

  It was entitled A Method of Reconstructing the Primary Alphabet from a Single One of the Series of Secondary Alphabets, and its 15 pages comprise Friedman’s first writing on cryptology. The primary alphabet is the mixed alphabet used to form a Vigenère-like tableau for polyalphabetic encipherment; the secondary alphabet is the one recovered by the cryptanalyst. For example, a primary alphabet based on the keyword ABOLISHMENT may be slid against itself like this

  so that plaintext a = N, b = T, o = C, and so on. The cryptanalyst, however, not knowing the order of the letters in the plaintext alphabet, will arrange them alphabetically in his recovery, thus obscuring the keyword in the lower, or secondary, alphabet. It will look like this (to use Friedman’s own example):

  Friedman showed that the original alphabet could be recovered by making a chain of letters and then stretching them out at trial intervals of 1,2, 3,… 25 letters. To make the chain, the cryptanalyst takes as its first link the letter under a, which is N. He then finds this letter in the upper alphabet and takes as the second link the letter beneath it, or Q. He finds q in the upper alphabet and takes as the third link the letter beneath it, B. After completing the chain, he writes the letters out in successive trials, leaving ever wider spaces between them until he can see plaintext fragments that might form part of the keyword. Usually he does not have to write out the full chain before feeling that an attempt is useless. In this case, likely sequences appear at an interval of 9—which is the displacement of the two alphabets. Completion of the chain at this interval will produce the original ABOLISHMENT alphabet:

  Such a determination can be of crucial importance. Knowing the primary alphabet will enable the cryptanalyst to solve much more easily cryptograms based on it but with different periodic keywords. He will also be able to solve much shorter cryptograms. This glimpse of the underlying key system may help him solve messages in other primary alphabets. The technique has many implications, and cryptanalysts must often be grateful to Friedman for devising it.

  Riverbank Publication No. 16, Methods for the Solution of Running-Key Ciphers, showed in its 42 pages how to crack polyalphabetic ciphers keyed with long texts to defeat Kasiski analyses. Friedman set up an abbreviated tableau in which only the high-frequency key letters and plaintext letters appeared with their cipher equivalents in the known alphabets. Given a cryptogram, “the first step is to assume that the key-text and plain-text consist solely” of those letters; the possible combinations that could yield the actual letters of the cryptogram are set out, and the cryptanalyst attempts to anagram so that he obtains intelligible text in both key and plain. He then extends the fragments thus obtained by working the one text against the other. No. 17, An Introduction to Methods for the Solution of Ciphers, was simply that. No. 18, Synoptic Tables for the Solution of Ciphers and A Bibliography of Cipher Literature, set out cipher systems in a tabular arrangement similar to that devised by Porta.

  No. 19 was a highly original attempt to mechanize the solution of transposition ciphers. The basic idea of Formulae for the Solution of Geometrical Transposition Ciphers was conceived by Lenox R. Lohr, then a captain taking one of the Riverbank courses and later president of Chicago’s Museum of Science and Industry. While the formulas worked perfectly and produced plaintext fragments without the cryptanalyst’s having tediously to write out innumerable trial transposition rectangles, the geometrical, or route, transpositions for which they were designed were so rarely used that the work had little practical value. Nevertheless, it was a forerunner of techniques used today with electronic computers. In No. 20, Several Machine Ciphers and Methods for Their Solution, Friedman amplified the de Viaris solution of the Bazeries cryptograph, to which he gave the generic name of a “multiplex” system, and devised a solution for the Wheatstone crypto
graph, perhaps based on his work with the Pletts device. No. 21, Methods for the Reconstruction of Primary Alphabets, written in collaboration with Mrs. Friedman, took up where No. 15 left off. It extended the method of that brochure to secondary alphabets that resulted from the interaction of two different mixed alphabets that had been used as the plain and cipher components.

  Riverbank Publication No. 22, written in 1920 when Friedman was 28, must be regarded as the most important single publication in cryptology. It took the science into a new world. Entitled The Index of Coincidence and Its Applications in Cryptography, it described the solution of two complicated cipher systems. Friedman, however, was less interested in proving their vulnerability than he was in using them as a vehicle for new methods of cryptanalysis. Fabyan had the pamphlet printed in France in 1922 to save money; General Cartier saw it and thought so highly of it that he had it translated and published forthwith—false-dating it “1921” to make it appear as if the French work had come first!

  In it, Friedman devised two new techniques. One was brilliant. It permitted him to reconstruct a primary cipher alphabet without having to guess at a single plaintext letter. But the other was profound. For the first time in cryptology, Friedman treated a frequency distribution as an entity, as a curve whose several points were causally related, not as just a collection of individual letters that happen to stand in a certain order for noncausal (historical) reasons, and to this curve he applied statistical concepts. The results can only be described as Promethean, for Friedman’s stroke of genius inspired the numerous, varied, and vital statistical tools that are indispensable to the cryptology of today.