by DAVID KAHN
Barnes was in close contact with Major Frank Moorman, chief of G.2 A.6, and it may have been Moorman who proposed that the A.E.F. use unenciphered two-part codes changed either before the Germans could solve them—a period estimated at from two to four weeks—or upon the capture of a book. On May 24, Hitt was writing to Moorman: “I concur in your ideas about the trench code book. I believe that we can republish it every two weeks….” Barnes, who was no cryptanalyst, acquiesced in the views of those who were. Frequent replacement was the principle of the Satzbuch, but the American codes were intended for service closer to the front. Thus the burden of augmenting security was lifted from the front-line soldiers and thrown, in the form of the more complicated two-part arrangement and the rapid replacement of codes, on the relatively undistracted personnel at headquarters.
On June 24, 1918, the Code Compilation Section published the first of the superb series of A.E.F. field codes—the Potomac. A 47-page booklet, it contained about 1,800 words and phrases for tactical needs (during the night = ANF, machine gun ammunition = APU). About 2,000 copies were printed and turned over to G-2 for distribution as far down as battalion headquarters. It went into service on July 15. The Potomac set the pattern for subsequent codes, which were printed and held in reserve, one set at army headquarters, a second at General Headquarters. Thus when, as expected, the Potomac Code was captured a few weeks after its publication, it took only two days to issue the Suwanee to the entire A.E.F. The Wabash moved into place as the back-up code and then, 16 days later, into service. It was followed by the Mohawk, Allegheny, Hudson, and Colorado codes at intervals of 3, 9, 21, and 22 days.
The rapid growth of the A.E.F. necessitated an increase in the number of copies printed to 3,200, but this also increased the danger of capture. So with the formation of the 2nd Army, a series of codes named for lakes was instituted on October 7 for the use of that army, while the river series was continued for the 1st Army. The Champlain, Huron, Osage, and Seneca codes were issued to the 2nd Army at intervals of 8, 13, and 9 days. At the Armistice, the Niagara Code was in press and the Michigan and Rio Grande codes in manuscript. In the five months between June and November, the section turned out nearly three codes a month—a noteworthy achievement, particularly in comparison with what the other belligerents accomplished.
Portions of the encoding and decoding sections of the A.E.F.’s Hudson Code
The Code Compilation Section printed its codes under conditions of tightest security at the Adjutant General’s printing office at Chaumont. Codes took priority over all other work except general orders and bulletins. Under favorable conditions, a field code would go from manuscript to binder in five or six days. Each code was proofread twice. “During the process of printing,” Barnes wrote, “the codes were under the constant supervision of an officer whose duty it was to destroy all spoiled sheets containing impressions even to the mats on the presses. All copies were counted and accounted for and the metal type melted down after the final impression. In many cases, two or three officers were on duty in the printing office keeping the various operations in sight.” The size of an edition was determined by G-2. Because the courier service refused to carry the heavy packages of codebooks, G.2 A.6, whose personnel realized the importance of secrecy in communications, took over the actual distribution of copies. Officers at the headquarters where the codes were kept in reserve were ordered to make frequent checks of the number of packages and the seals on them. A British officer was dumfounded when he heard that American codes could be prepared in ten days, saying that it would take his army at least a month.
Their cryptanalytic resistance was, as with the enciphered code, gauged by actual test. This time the results were positive. Members of G.2 A.6 reported that the system, while not insoluble, excelled that of the Germans. Coded messages had been sent to the British for further examination. Hay reported that “We have not been able to solve them or even to get any light. The security appears of a high order.” Hitchings wrote: “I am sending you a short survey of our observations on the 41 messages…. we have not succeeded in solving them, but you will see in the enclosed survey a few possible lines of attack.” And while Parker Hitt had not tried to solve any messages in the code, his general experience led him to say, “We believe that this code system will be better than anything now in use on either side.”
These field codes served primarily for communication within each division, though they also encoded messages between divisions and to higher headquarters. Battalions on the flanks of each army exchanged codebooks to permit intercommunication. The A.E.F. supplemented them with a variety of others needed by a modern army and prepared by the Code Compilation Section. Troops in the very first trenches used the Emergency Code List, a single sheet with about 50 common expressions represented in two-part arrangement by two-letter groups (CM = message not understood; PV =our artillery is shelling us). It resembled the carnet de chiffre. New editions were distributed at the same time as new editions of the field codes. For headquarters work Barnes’ section produced 1,000 copies of the massive Staff Code: 30,400 words and phrases, whose four-letter codewords, in one-part order, were superenciphered digraphically, with different tables for G-1, G-2, G-3, G-4, and G-5. It was probably the largest codebook ever printed in the field. There were also special codes for reporting casualties, for technical radio matters, for extra secrecy at six major telegraph posts in reporting troop movements, and for designating the names of organizations and officers over the telephone by using women’s names as jargon (28th Division = JENNIE: Chief of Staff = DOW; Chief of Staff of 28th Division = JENNIE DOW). In its ten months of active work, the section printed more than 80,000 codebooks and pamphlets, all numbered, recorded, issued and receipted for.
Front-line cryptography: an A.E.F. code list issued for use in the trenches
In addition to these official codes, many A.E.F. units cooked up their own unauthorized ones. In the 82nd Division, for example, officers said GREAT NECK for Grosreuves and BUZZARD for 1st Battalion, 326th Infantry. Some anonymous but avid baseball fan in the 52nd Infantry Brigade produced the gem of these unofficial systems. If we were under bombardment, it was WAGNER AT BAT; if the Germans simply lobbed over some enemy registration fire, WAGNER BUNTED; if we were under light bombardment, WAGNER DOUBLED, and if we were under heavy bombardment, WAGNER (whose nickname, it will be remembered, was “Hans”) KNOCKED A HOME RUN. Juvenile all this may be, but if codes are to delay enemy comprehension, this one no doubt served its purpose.
But the finest codes in the world, changed at the most rapid intervals, are worthless if wrongly used. Did the American doughboy fulfill his opportunity under these codes to achieve a superior security of communication? He did not. His irritation at the nuisance of encoding and his consequent unconcern for regulations could be matched against any combatant’s. Encoding delayed signaling, and combat officers bitterly resented this gumming of communications just when they were most needed. Aversion became so extreme that at one point a general actually gave his division specific orders to use no code before and during an important movement. The order was undoubtedly born of some unhappy experiences, and it was, in any case, less dangerous than any semicoding or other violations of coding regulations, such as sending messages to addressees not having the code, necessitating repeats either in clear or another system. The well-known American disregard for regulations—especially ones as persnickety as these—and the tendency to take the easiest way out caused G.2 A.6 chief Moorman to remark exasperatedly that “there certainly never existed on the western front a force more negligent in the use of their own code than was the American Army.”
Violations, in fact, became so numerous that a Security Service was set up to monitor American radio messages (later, telephone conversations as well). Its first station began operating at Toul on July 11, 1918; eventually the A.E.F. had four. The messages were sent to an officer in G.2 A.6 who studied them for practices that would help the Germans in solution. Letters pointed t
hese faults out to commanders. One sent by the adjutant general to the commanding general of the 1st Army relating to a single message of September 17, pointed out that the plaintext Boche, spelled out by five code groups, could have been replaced by German or enemy, each a single codegroup, that two groups for day light could have been used instead of the 18 for almost before the crack of dawn, that work could have been written instead of business with a saving of seven groups, and so on.
Most of these rather fussy letters were ignored. “Only a few of these were answered,” Moorman complained, “and in these cases the action taken was entirely inadequate. In one case an officer was reprimanded by his commander. In others the excuse was made that officers did not know or were too busy or thought they were justified in their action…. in trying to check up and eliminate faults we have found great willingness and ability to refer us to someone else.” He proffered a unique solution of his own to end the vexing problem: “My idea would be to hang a few of the offenders. This would not only get rid of some but would discourage the development of others. It would be a saving of lives to do it.” Barnes’ more moderate idea of assigning a cryptographic control officer to each headquarters was preferred, but it did not go into practice until 20 years later.
Perhaps the most interesting thing about the entire American cryptographic operation was the attitude taken toward it by Barnes and his men. They did not regard their codes as immutable; rather they sought continually to improve them. Further, their efforts encompassed the physical as well as the cryptographic aspects. Paper, for example, was chosen so that it would stand up just long enough for the brief life of the book and would burn easily in case of danger. The typeface—named “Typewriter”—was picked for its legibility in the ill-lit dugouts of the front. The books continually shrank in size from the 7¼ × 9¾ inches of the Potomac to the 5½ × 7½ of the Colorado and subsequent books. In the later books, nulls were prominently bunched next to the encoding columns to encourage their use, and common suffixes, such as -ing, were listed conveniently at the bottom of each page. Homophones grew more abundant, and blanks were provided for special terms or names needed within the different divisions. A G-2 circular inviting suggestions brought in many requests to include certain phrases. To use them all would have swollen the book beyond easily manageable proportions, and Barnes winnowed out the many local and transitory ones. But the adaptability of the Code Compiling Section is shown by the fact that almost half of the 1,900 words and phrases in the Osage Code were new compared to those in the Potomac.
The section never satisfactorily resolved a continuing dispute over the relative merits of letters or numbers as codegroups, though it consulted many telegraphists, radio operators, code clerks, and experienced code officers. Opinion was almost equally divided. Most of the codes used three-letter codegroups, but a few were published with four-digit ones in an apparent experiment to see which actually worked best. The same undogmatic approach was demonstrated in the submission of the books for cryptanalytic tests, and in the testing of 50,000 telegraphic combinations to empirically select those resulting in the fewest errors as codegroups for the Staff Code.
In short, the Code Compiling Section was willing to learn, and it did learn a great deal that notably improved American codes. To an astonishing degree, it encapsulated “that practical, inventive turn of mind, quick to find expedients,” that historian Frederick Jackson Turner found the frontier had shaped as an American trait. Perhaps this is best epitomized—with the important addition of some American humor—by the codegroup to report that the code had been lost. The early codes did not even have one. The Hudson Code displayed in large type on its cover, “Memorize this Group: ‘2222—Code Lost.’ ” Then the codegroup for Code Lost was changed to DAM.
To the right of the imposing dark stone headquarters building at Chaumont stood an undistinguished, single-story barracks of glass and concrete. Sometimes called the “Glass House,” the caserne housed the other half of the American cryptologic effort, the Radio Intelligence Section, G.2 A.6.
Its chief, Moorman, 40, a native of Greenville, Michigan, was a blue-eyed, brown-haired Regular Army man who had worked his way up through the infantry ranks from private. He was a 1915 graduate of the Army Signal School and knew enough about cryptanalysis to devise an ingenious method for almost automatically determining the letters of a Playfair keyword. Hitt thought it valuable enough to include in his Manual. In France, however, Moorman did not engage in any actual cryptanalysis, except perhaps to help out, since his work as head of G.2 A.6 was administrative, not operative. As a boss he was well regarded by his men for his fairness and blunt honesty.
His organization began to take rudimentary shape in the fall of 1917 with a mere handful of men, the nucleus of what became a 72-man unit at the period of the A.E.F.’s maximum expansion. They came from the most varied civilian occupations. There were two New York lawyers, both lieutenants—Hugo A. Berthold, who was of Germanic extraction, knew the language well, and became Moorman’s chief assistant and head of code cryptanalysis, and Robert Gilmore. Childs, who had solved the superencipherment, had been a reporter on the Baltimore American before taking his M.A. at Harvard in 1915. Lieutenant Lee West Sellers was a New York music critic, and Lieutenant John Graham an instructor at Washington and Lee University, later a professor of Romance languages there. There was an architect who had studied Hebrew, Persian, and other Oriental tongues; one man was a chess expert, another an amateur archaeologist. About the only two who had had any experience at all with codes or ciphers were Corporal Joseph P. Nathan, who had worked in the code section of the Grace Line in New York, and one not unknown to later fame, Lieutenant William F. Friedman, who had become interested in the subject several years earlier. In addition to these men, six cryptanalysts were assigned to each army headquarters to decrypt intercepts from their front with keys from G.H.Q.
The work of G.2 A.6 divided into cryptanalysis and four minor areas—traffic analysis, intercepting enemy telephone conversations, following enemy air artillery spotters, and checking monitored American communications for security breaches. These minor functions made more important contributions than it would at first seem. Moorman, for example, originally did not consider the traffic analysis particularly necessary. But he saw its value when his men became skilled enough to draw a map of the German order of battle and to see through German fake messages. They even managed to discover two newly formed armies and thus help give warning of a new German drive. The aircraft teams eavesdropped on the planes as they signaled targets to their batteries and warned Allied troops that they were about to be fired upon; sometimes the G.2 A.6 experts even identified the battery that was about to fire, permitting Allied counterbatteries to shell them first.
The monitoring officer, Lieutenant Woellner, came up with some frightening object lessons. He deduced the entire American order of battle for the assault on the Saint-Mihiel salient from monitored telephone messages, missing the time of attack by 24 hours only because one speaker had misstated it! Most of his information came from a single switchboard operator who complained that certain lines had been broken by tanks and heavy artillery moving into a small woods near him all night. “Whether or not the Germans picked up this message we never learned,” Moorman commented disgustedly, “but it is certain that this one operator did all that could be reasonably expected of one man in the matter of telling the Germans when and where the attack would take place and the forces to be engaged.”
Like the other sections, the cryptanalysts got off to a slow start. Their training had been all in ciphers, whereas the Germans were using code. In November of 1917, Berthold went to the French cipher bureau, where he picked up some instruction and probably some current KRU solutions as well. With this help, G.2 A.6 discovered that certain nearby German stations radioed regular reports at regular hours—a habit that thenceforth proved fatal to many a Satzbuch. By the end of the first week of a code’s month-long life, Moorman said, “we were reading some
of the routine messages…. At the end of the second week we were reading many of the messages, and at the end of the third week we practically controlled the code. This really meant that we had for one week a real control of each code.”
G.2 A.6’s first real victory in the war of the intercepts came with the introduction of the Schlüsselheft. The success was due in large measure to the alertness of the Signal Corps’ Radio Section, which operated the network of intercept stations that fed the cryptanalysts their raw material. The first stations were set up in the fall of 1917, and by the end of the war the five posts had snatched 72,688 German messages from the airwaves. Eight direction-finding stations took the astonishing total of 176,913 bearings. The radio operators, frequently working in damp and drafty shacks exposed to enemy fire, won high praise from the cryptanalysts for the accuracy of their interception of long strings of meaningless letters. Often they picked up messages that the other Allies had not heard, and this was what happened on March 11, 1918.
It was at midnight of that date that the Germans placed into service not merely a new code, but one that, from its numerical codegroups, appeared to be of a different breed entirely. The Allies were expecting a major German push, and the appearance of this code was considered another straw in the wind. Its solution would obviously be of importance in giving clues to German activities. Though the British had suggested that a superencipherment might be involved, the precise nature of the system had to be determined, the superencipherment stripped off, and the repertory then built up. This would have imposed much greater difficulties than just solving another Satzbuch edition—except for American alertness.
