THE CODEBREAKERS

by DAVID KAHN

  As a result, Austrian cryptanalyses declined sharply in the latter half of the war. Nevertheless, Austria-Hungary had enjoyed the preponderance of cryptanalytic success on the Southern Front. Ronge always cherished as the greatest tribute to his Dechiffrierdienst an unintended one from the foe. A postwar commission of enquiry into the Caporetto disaster reported with anguish that “The enemy had known and deciphered all our codes, even the most difficult and most secret.”

  11

  A WAR OF INTERCEPTS: II

  NINETEEN ELEVEN is not a momentous year in American history. The last two territories on the continent, New Mexico and Arizona, were preparing for admission to the Union. The large-girthed William Howard Taft lumbered about the White House, trying to ignore the pyrotechnics of his predecessor, Theodore Roosevelt. C. P. Rodgers made the first airplane flight across the country. Carry Nation died. Perhaps the most impressive event of the twelvemonth was Ty Cobb’s batting that incredible .420. The year was not outstanding, but it was the year in which the United States took its first faltering steps in official military cryptanalysis.

  They were taken at Fort Leavenworth, Kansas. Here America’s tiny prewar Army had its Signal School. In 1911, the school began a series of technical conferences, and on December 20 portions of a paper on “Military Cryptography” by Captain Murray Muirhead of Britain’s Royal Field Artillery were read to Conference No. 4. The students responded with some papers of their own. Captain Alvin C. Voris showed how unsuitable the purely administrative War Department Telegraph Code was for troops in the field and proposed a tactical supplement for it. Lieutenant Frederick F. Black praiseworthily made an attempt to mechanize en- and deciphering by putting caps over typewriter keys. Lieutenant Karl Truesdell took a basic first step by compiling 10,000-letter frequency tables for English, German, French, Italian, Spanish, and Portuguese. A few months later, Lieutenant Joseph O. Mauborgne—who was to become Chief Signal Officer—whiled away the long hours of a trans-Pacific crossing by solving an 814-letter Playfair from Muirhead; he described his methods in 1914 in a 19-page pamphlet that is the first published solution of that cipher.

  The Muirhead seed ripened best in the fertile mind of a 34-year-old captain of infantry named Parker Hitt. Hitt was the towering figure of American cryptology in those days, both figuratively and literally. Six feet four inches tall, a native of Indianapolis, he had left his studies in civil engineering at Purdue University in 1898 to join the Army. He served in Cuba, won a commission, and saw, if not the world, at least the Philippines, Alaska, and California. After graduating from the Signal School, he stayed on as an instructor. Hitt participated in the technical conferences and, among other things, demonstrated the insecurity of Black’s typewriter method by taking only 45 minutes to solve one of the automatic cryptograms.

  He discovered that he was “very much interested in cipher work of all kinds” and that he had a real knack for it. When the border command began intercepting Mexican cipher messages as American friction grew with that troubled country, the messages found their way to Hitt. Soon he was solving transposition ciphers, monalphabetics, polyalphabetics (some with mixed alphabets) used by agents of Pancho Villa and others, and a homophonic substitution used by the Constitutionalists. This had four numerical cipher alphabets, all of which remained fixed during the encipherment of a single message, but whose positions were changed from one message to another. The key could be indicated by the letters above the lowest number in each alphabet, or by the four numbers under A. For example, the arrangement used for a message between Saltillo and Juarez, intercepted on November 26, 1916, was:

  Hitt solved this, and many like it. The system later became more widely known under the name of the Mexican Army Cipher Disk when the four numerical alphabets were placed on revolving disks.

  Hitt demonstrated his acuity in cryptanalysis nowhere more strikingly than with a subtle numerical system forwarded him by Lieutenant Colonel Samuel Reber of the Office of the Chief Signal Officer. Reber wrote him on September 21, 1915: “Some time ago while in conversation with the Assistant Chief Engineer of the Western Electric Company, I told him that a good cipher expert could work out almost any cipher, and his letter of August 3rd shows what he thinks in the matter. I am sending you the ciphers….” On the 24th, Hitt, then at the School of Musketry at Fort Sill, Oklahoma, received the cryptograms, which were two strings of unbroken numbers, and the next day, a rainy Saturday, analyzed them. That afternoon he wrote Reber:

  “No. 1 consisted of 415 figures and the factors of this are 83 × 5. This led to the conclusion that I had five figure groups to deal with and this was checked affirmatively when I made out a list of these groups and found some duplicates and a few triplicates. The ratio of occurrence of these duplicates and triplicates led me at once to the conclusion that each group represented two let

  “The groups ran in value from 00518 to 53339 with large gaps. I then the small graph of group values and found that I could roughly superim normal frequency table on the graph, but the scale, if I may so call larger at the A end than at the Z end. This suggested a logarithmic sca reached for a table of logarithms.

  “00518 showed up as log 1012 and 53339 as log 3415 exactly. If A = 10, then 12 = C, 34 = Y and 15 = F. The rest of the solution merely involved the use of the logarithm table on these five figure groups and the reduction of the numerals so found to letters” In a few swift slashes of his mind he thus cracked an ingenious two-step cipher, to Reber’s pleasure (though he ungraciously said he could have done it himself if he tried) and to the chagrin of the Western Electric assistant chief engineer.

  During 1915 Hitt was working on a project that he had mentioned in a letter to Reber on January 15: “I have a mass of material on cipher work, the accumulation of the last four years, and hope to put it into shape as a pamphlet before I leave here if time permits. Major Wildman has kindly suggested that I do this in order that the pamphlet be used as a basis for the course in cipher work.” He enriched his own experience—greater than that of any other person in the country at that time—with theory and new information from European books on cryptology that he borrowed from the Army War College. He finally completed his booklet late in 1915, and the next year the Press of the Army Service Schools at Fort Leavenworth published 4,000 copies of his Manual for the Solution of Military Ciphers, selling it at 35 cents the copy.

  It was an excellent work. It naturally explained how to solve the standard ciphers, up to periodic polyalphabetics with mixed alphabets and—for perhaps the first time in the literature of cryptology—combined transposition-substitution. But its special merit lay in its practical tone. The book was imbued with a verisimilitude, an air of this-is-how-things-really-are, that stemmed largely from Hitt’s grounding in the realities of signal communication. This pragmatic approach cropped up, for example, in the book’s discussions of why cryptanalytic offices should be attached to field headquarters and how they should be organized, of the need for accurate intercept and recording procedures and how they may be achieved, and of how to correct errors in enciphering and transmission—a subject of the utmost practical importance and one almost invariably neglected in treatises. Hitt replaced the waxen examples of other books with real cryptograms, several with Spanish plaintexts, whose presence, in view of the Pershing punitive expedition, intensified the feeling of reality. As a military man, Hitt wrote with directness; as one with an extra measure of intelligence, he wrote with clarity; and as one with a touch of the poet, he flavored his 101 pages with a prairie tang all his own. “As to luck,” he observed when discussing the fourth of four factors that determine success in cryptanalysis (the others being perseverance, careful analysis and intuition), “there is the old miner’s proverb: ‘Gold is where you find it.’ ”

  Yet the book was outdated at the moment of its birth. Events in Europe had far outrun its elementary notions. Cryptograms were no longer being solved on the basis of single messages, as in Hitt’s examples. Military ciphers had long since
attained a complexity never hinted at in the Manual The French had anticipated his ideas on cryptanalytic organizations. The Spanish-language examples might better have been German. And in view of the trench codes which were then emerging as the dominant form of cryptography, one sentence was singularly inapt: “The necessity for exact expression of ideas practically excludes the use of codes for military work although,” he hedged, “it is possible that a special tactical code might be useful for preparation of tactical orders.”

  All this is true. Yet it remains equally true that the book filled a real need. Many people, struck by the interest in these matters that war always enlarges, wanted to know about cryptology. But the United States was achingly devoid of information: Hitt’s was, in surprising fact, the first book on the subject published in America*—and indeed the first devoted to cryptanalysis in English since Philip Thicknesse’s 1772 A Treatise on the Art of Decyphering! Soldiers and civilians grabbed at it. A second edition became necessary, and this time 16,000 paperbound copies were run off, giving it a greater circulation than any previous book in the history of cryptology. Elementary it may have been, but for those who knew nothing of the subject, a basic work was what was needed. When the United States declared war, Hitt’s Manual served as the textbook to train future cryptanalysts of the American Expeditionary Forces. Some of this training was done at the Army War College in Washington under the auspices of MI-8, the cryptologic section (number 8) of the Military Intelligence Division, headed by Herbert O. Yardley, and some at the Riverbank Laboratories in Geneva, Illinois, where cryptologic research, mainly aimed at proving that Bacon wrote Shakespeare, had been carried on since before the war. Riverbank also had some texts of its own.

  In doing the research for his book, Hitt ran across a military cipher that greatly impressed him as affording more security than any other that he knew. He, and probably all the other young cryptanalysts at the Signal School, stood aghast at what was then the “official” U.S. Army field cipher. This was the Signal Corps cipher disk, a celluloid device with a reversed cipher alphabet revolving inside a standard plaintext alphabet. The Army used it with a repeating keyword to produce a straight periodic Beaufort cipher. It was equivalent to the Confederate cipher disk of 50 years before and inferior to the cipher disk described by Porta three centuries before that—a record of retrogression unmatched, perhaps, by any science in the world. Even though the cipher disk was the “official” system, Hitt’s own 2nd Division used a then-popular cipher called the “Larrabee.” It was simply an ordinary Vigenère printed so that the plaintext alphabet was repeated for all 26 cipher alphabets. Neither it nor the cipher disk would have delayed an expert cryptanalyst for more than an hour. On May 19, 1914, Hitt had recommended that the Larrabee be replaced by the Playfair as the 2nd Division cipher, but was turned down. Undaunted, he proposed the new cipher that impressed him so much to the director of the Army Signal School on December 19, 1914.

  “This device is based, to a certain extent, on the ideas of Commandant Bazeries, of the French Army,” he wrote in his memorandum. Hitt in effect peeled the alphabets off the disks of the Bazeries cylinder and stretched them out in strip form. He cut 25 long slips of paper, printed a mixed alphabet on each of them twice, numbered them, and then arranged them in a holder in the order given by a keynumber. To encipher, he slid the slips up or down until they spelled out the first 20 letters of the message in a horizontal line, and then selected any other line, or generatrix, as the ciphertext, repeating this process until the entire message was enciphered. Hitt’s first holder was 7 × 3¼ inches. He also made the device in its original Jefferson-Bazeries form by sawing disks off a cylinder of apple wood.

  He requested that the device be forwarded to the Chief Signal Officer. About 1917, his old fellow student at the Signal School, Joseph Mauborgne, then in charge of the Signal Corps Engineering and Research Division, fixed the device in the cylindrical form for the Army and mixed the alphabets much more thoroughly than Hitt had, thereby making solution more difficult. In 1922, the Army issued its M-94, which strung 25 aluminum disks the size of a silver dollar on a spindle 4¼ inches long. The M-94 remained in Army service until early in World War II. Between the wars, both the Coast Guard and the Radio Intelligence Division of the Federal Communications Commission made use of it. In the 1930s, the Army reverted to Hitt’s slide form in its cipher device M-138-A, which improved on the Hitt device by providing 100 slides, 30 of which were used at a time. The State Department adopted the M-138-A in the late 1930s and early 1940s as its most secret method of communication. The Navy likewise used it very widely in World War II. It was commonly called the “strip system.” Thus Hitt’s few paper slides became one of the most widely used systems in the history of American cryptography.

  In 1917, Hitt went to France with Pershing’s staff as assistant to the Chief Signal Officer. When the A.E.F’s 1st Army was formed, Hitt became its Chief Signal Officer. Though there was no cryptology involved in this job, his book had made him the American expert on the subject, and his advice was often sought. It was even followed, since Hitt was widely respected.

  While he was overseas, his wife, Genevieve, who was operating the code room at Fort Sam Houston in San Antonio, struck up a friendship with a young lieutenant and his wife who lived across the way. Their names were Dwight and Mamie Eisenhower, and the friendship of the two families stretched across the years. One morning during World War II the Hitts stumbled across Ike, stretched out asleep in the living room of their home in Front Royal, Virginia, and in the 1950s Parker Hitt attended one of the famous and exclusive stag dinners given by the President at the White House.

  When the United States entered the war, its Army had no official codemaking or codebreaking agency. Codes were occasionally compiled, of course, and each unit seemed to prescribe its own field ciphers, as Hitt found out when he tried to replace the Larrabee with the Playfair. Any cryptanalysis was on a strictly informal basis, like the messages that were sent to Hitt, often with a request like that from the acting intelligence officer of the Southern Department on March 7, 1917: “1. The inclosed cipher messages have been received from the Chief of the War College Division, General Staff. 2. It is requested that you decipher them as they are unable to do it in Washington. 3. The results obtained are desired at the earliest practicable date.” (Hitt returned these on March 10 saying that they appeared to be in code and that he could not read them.) Usually Hitt had to squeeze this work in among his regular duties. The Riverbank Laboratories also did some informal cryptanalysis for the War Department.

  It was obvious, upon the arrival of the first token units of the American Expeditionary Force in France in the spring of 1917, that the A.E.F. would have both cryptographic and cryptanalytic work to do. Consequently, General Orders No. 8 of July 5, 1917, which established the A.E.F. headquarters organization, provided for these functions. It assigned “American codes and ciphers” to the Signal Corps but gave “policy regarding preparation and issue of ciphers and trench codes” to the Intelligence Division, probably because this was also charged with “enemy’s wireless and ciphers” and “examining of enemy’s ciphers.” Having the cryptanalysts supervise the cryptographers was excellent in theory—and it worked out fine in practice. The two organizations that came into being in accordance with this order collaborated closely throughout the war. One was G.2 A.6, the Radio Intelligence Section (the 6) of the Military Information Division (the A) of the Intelligence Section (the 2) of the General Staff (the G). The other was the Code Compilation Section of the Signal Corps. Both were stationed at American G.H.Q. at Chaumont, a town on the Marne about 150 miles east of Paris.

  The Code Compilation Section was set up in December of 1917. There had been no real need for it before then because the United States had no troops in the line. In command was Howard R. Barnes, a 40-year-old Ohioan who had been commissioned a captain because of his ten years of experience in the State Department code room. Under him were three lieutenants and a corpor
al. The unit examined and discarded the three means of secret communications then authorized for the A.E.F.—the War Department Telegraph Code, which, as Voris had pointed out, was unsuited to tactical work, the cipher disk, whose security was nil, and the Playfair, which could not sustain security under regular use, but could and did serve as an emergency system.

  Cryptography on the Western Front had evolved through ciphers to codes, and Barnes, bowing to this experience, began the task—never before attempted in the American Army—of compiling a codebook in the field. His section studied an obsolete trench code that the British had reluctantly turned over, made firsthand observations of communication needs at the front, and drew up The American Trench Code, of 1,600 elements, and the Front-Line Code, of 500. Both were one-part codes to be used with a monoalphabetic superencipherment. The 1,000 copies of the Trench Code were distributed only down to regimental headquarters, and the 3,000 copies of the Front-Line Code to companies. They served as the American cryptosystems during the first weeks of real A.E.F. participation in the war—the weeks of Château-Thierry and Belleau Wood.

  But the system of enciphered code did not last long. Barnes frequently consulted with Hitt—“To him more than to any other officer of the American Army is due whatever success the American Codes may have obtained,” he later wrote—and Hitt suggested testing the superencipherment. G.2 A.6 lent Lieutenant J. Rives Childs. On May 17, 1918, Childs was given a copy of the codebook and 44 superenciphered messages. Within five hours—three of them spent just in making frequency counts—he had recovered the encipherment alphabet. At about the same time, Barnes and his men realized that a superencipherment imposed extra delay and extra work upon the encoders at the front, with all the dangers that that entailed. Superenciphered code would have to be junked. But what would the A.E.F. use?