THE CODEBREAKERS

by DAVID KAHN

  German victories drove the French back into their own territory, where they used their own wire network for communication and thus deprived the enemy of much chance of intercepting radio messages. The same situation freed the French radio for intercept work whereas the Germans had to use their wireless for communication. French cryptanalysis thus owed much of its success to the highly dubious advantage of having the war fought on French territory. One may wonder whether the French would have preferred solving enemy cryptograms or the nondesolation of dozens of villages, orchards, fields, and forests in their northern provinces.

  As the war progressed, the French began using radio more and more. By 1916, the Germans awoke to their opportunities and set up the Abhorchdienst (“Intercept Service”). Its main station was at Neumünster, where cryptanalysts, many of them recruited from the ranks of mathematicians, were soon solving Playfairs within a day after a key change. Later, the Germans established a cryptanalytic center at their Western Front G.H.Q. at the Belgian resort of Spa. But they never caught up with the Allies, who had had the inestimable advantage of familiarity with German phraseology and idiosyncrasies, gained in the first chaotic days, and of preventative improvements in their own communications.

  Both sides, however, were equally adept at picking up the enemy’s frontline telephone messages—an eavesdropping that was facilitated by the fixed nature of trench warfare. Conversations could be heard either by induction through earth pickups, or by actual taps of enemy wires by intrepid soldiers who crawled across no-man’s-land. Both sides obtained enormous quantities of intelligence from this source. Officers and men repeatedly violated the strict regulations against transmitting any important information over field telephones.

  In 1916, for example, the British sustained casualties in the thousands in a fierce battle to take Ovillers-la-Boiselle on the Somme. Battalions were decimated as they went over the top. When the British finally captured their objective, they found in one enemy dugout a complete transcript of one of their operation orders. A brigade major had read it in full over a field telephone despite the protest of his subordinate that the procedure was dangerous. “Hundreds of brave men perished,” the British signal historian related, “hundreds more were maimed for life as the result of this one act of incredible foolishness.” The search for protection resulted in the ultimate cryptographic development of the First World War. These were the trench codes.

  In February of 1916, General Auguste Dubail, the handsome and energetic commander of the French Army of Lorraine, requested some kind of code for telephone use because indiscretions had drawn so many heavy bombardments onto his reserves. The cryptographic office produced a carnet de chiffre (“cipher notebook”). Important words in telephone messages were to be spelled out in code form by replacing their letters with the two-digit groups of the carnet. Soon a table of 50 common expressions was added, and the carnet authorized for use by wireless telegraphy. This spurred its enlargement into a small code of three-letter groups for use by smaller units. This was called a “carnet réduit” (“condensed notebook”) in contrast to the larger headquarters codes.

  The carnets were replaced from time to time. Each had a name—OLIVE, URBAIN, and so on—and the initial letter of that name, repeated three times, indicated the carnet that had encoded the message. The carnets were caption codes: the plaintext elements were arranged in categories, such as artillery, infantry, numbers, letters, common words, prepared phrases, place-names, verbs, and so forth. Though the codewords of the early carnets ran in alphabetical order, the topical distribution of the plaintext ruffled the one-part aspect of the code. Later carnets thoroughly mixed the codewords as well.

  Germany did not start using codes until a year after France did, but then they evolved in roughly the same way.

  The simple Befehlstafel (“command table”) came first. A small trench code in which bigrams represented common words or letters, it superseded the grilles in March of 1917. Some Befehlstafeln were in the form of notebooks with variable pagination; others were constructed as cipher disks, in which a change of position would give a change of equivalencies. In June these were supplemented on the regimental level by the Satzbuch (“sentence book”), the German version of the French code chiffré. The 2,000 (later 4,000) plaintext expressions of the Satzbuch were represented by thoroughly mixed three-letter codewords. It provided numerous homophones (anschluss fehlt, [“link-up missed”] = KXL, ROQ, UDZ) and many Blinde Signale, or nulls. Unlike the code chiffré, it was not superenciphered; it relied instead on planned obsolescence for security. At first a new codebook was issued about every month, but the interval was gradually cut down to about 15 days. This multiplicity of codes in time was matched by one in space. Where at first the entire front shared a single code, soon army groups and then individual armies had their own Satzbucher.

  The French called these codes the “KRU” or “KRUSA” codes, because all their codewords began with one of those five letters. The first one disconcerted the Service du Chiffre, unaccustomed as it was to dealing with two-part German codes. But it recovered quickly and, with Déjardin playing a leading role, reconstituted it sufficiently to read most messages. As the number of codes multiplied, their successful solution depended increasingly on accurate traffic analysis—an accurate separation of the messages of one army from those of another. This was managed, and the French soon were straining to recover the first 100 or 150 groups of each code as quickly as possible, for with this entry the rapid filling out of the repertory was virtually assured. Most of the 30 German codes that France solved during the war must have been Satzbucher. The information obtained during the ten days from December 5 to 15, 1917, a period picked at random, illustrates the value of the cryptanalysis: discovery of four division movements, reconfirmation of the identity of 32 regiments, ascertainment of the presence of a counterattack division north of St. Quentin, and warning of a German surprise attack at the Abia farm, which the alerted French troops repulsed.

  In March of 1918, the British predicted that the Germans would soon change their trench codes, probably in the direction of enciphered code. Painvin and Cartier were discussing this possibility with a visitor when Painvin was called to the telephone. French G.H.Q. informed him that what appeared to be that very switch had been made that day over the entire front, replacing the Befehlstafel trench code. The basis of the new system was the Schlüsselheft (“keybook”), a caption code of 1,000 three-digit groups. Only the first two digits of each codegroup were enciphered. This was done with a Geheimklappe (“secret flyleaf”), a 10 × 10 table with placode digits 0 to 9 as coordinates on the top and side and the encicode digits dispersed irregularly inside. Toward the end of the war, the Geheimklappe changed daily.

  Though the cruel deadlock of the Western Front riveted the major attentions of the Entente and of Germany, its chief antagonist, battles on the Eastern and the Southern Fronts sacrificed their millions as well to the clash of national ambitions. Russia, isolated by the cruise of the Göben, hurled her mighty forces against the German and Austro-Hungarian empires time and again in noble resolution of her treaty obligations; her eventual downfall, in no small degree a matter of cryptology, is a story in itself. In May of 1915, Italy denounced its treaty with the Central Powers and joined the Allies; Rumania followed a year later. Bulgaria lined up with Germany; Greece and Portugal with the Entente. Fighting blasted the Holy Land. All Europe and the Near East flamed.

  Thanks to its prewar training, the Austro-Hungarian Army’s Dechif-frierdienst handily unwrapped the Russian systems, aided by the innumerable confusions of mobilization. They had gained almost a year of invaluable wartime experience by the time hostilities broke out with Italy. Thus they achieved their first solutions of Italian cryptograms (of no tactical importance) on June 5, 1915, only 13 days after the declaration of war. These first four were followed by 16 others in June, most intercepted by the new station erected at Marburg. On July 5, the Austrians picked up their first dispatch in the cifrario
rosso (“red cipher”), the Italian staff cipher, which intelligence chief Ronge had prudently acquired before the war. They had the odd pleasure of reading a reprimand from General Luigi Cadorna, the Italian commander in chief, to Lieutenant General Frugoni for not having pressed an attack vigorously enough.

  Five days later, the cifrario rosso key changed. The Italian specialists among the Austrian cryptanalysts, spearheaded by the chief of the entire cryptanalytic section, Major Andreas Figl, cracked it only after considerable work. The number of solutions fell to 13 in July. But as the Austrians accustomed themselves to Italian methods, their successes waxed. By August 12, they had read 63 messages and could send the new key to the several army headquarters, where Figl had just stationed cryptanalysts. Captain Albert de Carlo was assigned to Bozen; Lieutenant Alfred, Baron von Chiari, went to the 11th Army at Adelsburg in the Tyrol and Lieutenant Hugo Scheuble to the 10th Army at Villach in Carinthia. Soon afterwards the Austrians captured the enemy’s field radio instructions, and thereupon the number of solutions mounted to 50 and sometimes 70 a day. Though these usually contained only administrative matters, they enabled Colonel Ronge to predict the course of impending offensives.

  By now the Austrian cryptanalysts had become so expert that they hardly noticed the changing every six weeks of the key of the field cipher, the cifrario servizio (“service cipher”). In October, the Italians put into front-line service a new system, the cifrario tascabile (“pocket cipher”), and Ronge boasted that “it was another one of my peacetime purchases that was already paying for itself.” For once Ronge was wrong: it had been a complete waste of money. The cifrario tascabile was no more or less than a Vigenère with the digits 1 to 0 tacked on to the end of the plaintext alphabet and with cipher alphabets consisting of the digits 10 to 45 in normal order ! Passwords usually served as keys. It should have taken the experienced Austrian cryptanalysts perhaps three or four hours at the most to identify and solve the first message or two in the system.

  This system was the brainchild of Felice de Chaurand de Saint-Eustache, an Italian colonel who before the war had laboriously solved a correspondence carried on alternately in two enciphered commercial codes, the Sittler and the Mengarini. Subsequently he “enhanced” his cryptologic reputation by devising the cifrario tascabile. It should have rather brutally exposed his ignorance—and it reflects badly on the poverty of prewar Italian cryptology that it did not. Anyone having the slightest acquaintance with the field would have seen the vulnerability of the cifrario tascabile, while anyone who had kept up with the literature would have known that de Chaurand could have solved his code correspondence in a few hours if he had applied Valério’s mechanical technique instead of requiring the two months of several hours’ work a day that he said, rather pridefully, it took him. Later, for some inexplicable reason, an Italian expeditionary force in Albania corresponded in this very same Mengarini code!

  During the big Austrian drive in the spring of 1916, Austrian cryptanalysts preyed not only on the cifrario tascabile, which was an easy killing, but on the other systems as well. One radiogram was intercepted during the evening of May 20; by 3 the next morning Figl’s group had read of arrangements for a heavy counterattack with reserves; by 4 countermeasures had been ordered which checked the Italian onslaught. On June 1, the armies’ intercept-cryptanalytic posts—which Ronge had codenamed “Penkalas,” after a pencil factory’s trademark that showed a head with a mechanical pencil behind an oversized ear—detected a change in Italian call-signs and cipher key. Four days later, a new call-sign was heard which later proved to be that of a newly formed Italian 5th Army. On June 8, the Italian 1st Army cipher key changed, and the air force got its own code. The Nachrichtenabteilung put these indications all together and they spelled “attack.” Consequently the Austrians were prepared for the Italians’ summer offensives on the Isonzo River. The cryptanalysts soon became so expert that the now-daily Italian key changes caused less trouble to them than to the legitimate decipherers. And when a new system was introduced on August 20, they cracked it within 38 hours.

  Cryptanalysis had thus become one of the major sources of Austrian intelligence, and by April of 1917, the organization that generated this information had burgeoned into a multisection outfit. Attached to the general staff’s Evidenzgruppe were Chiffrengruppe I, under Captain de Carlo, and Chiffrengruppe II, under Captain Richard Imme. Theoretically under the Evidenzgruppe, but, according to Colonel Ronge (who commanded both), the “real” Austrian intelligence service, was the Nachrichtenabteilung of G.H.Q. at Baden. One of its five divisions was the Kriegschiffregruppe (“War Cipher Group”), headed by First Lieutenant Hermann Pokorny, a brilliant cryptanalyst who had solved the first Russian cryptogram of the war, and who later became chief of the Evidenzgruppe. The Kriegschiffregruppe had three sections: an Italian under Major Figl (who later rose to colonel), a Rumanian under Captain Kornelius Savu, and a Russian under Captain Viktor von Marchesetti. Feeding intercepts to them were three major Penkalas: Austro-West, covering the Italian sector; Austro-Sud, the Rumanian; and Austro-Nord, the Russian. The entire complex was referred to by the unofficial title “Dechiffrierdienst.”

  Savu’s group, incidentally, made little progress for a while after Rumania’s entry into the war in 1916, but then the ciphers caved in and proved a mine of information, giving the Austrians full warning, for example, of a planned counterattack on September 14. Captain Franz Jansa, in charge of Austro-Sud, and his assistant, Captain Konstantin Marosan, found themselves so overworked that a cryptanalyst had to be attached to 1st Army headquarters. Later the flood slackened, but on occasion the Austrians read messages that the intended recipients could not, showing that they had not lost their touch.

  They did not capture all the laurels, however. Italy had made no prewar cryptographic purchases, but she was aided in her efforts to catch up to her enemy by some remarkably inept Austrian cryptography and some remarkably able Italian cryptanalysts.

  The first and best of these was Luigi Sacco, an enthusiastic, 32-year-old lieutenant of engineers at the Supreme Command’s radio station. He had first become interested in cryptography in 1911, at the time of Italy’s war with Turkey. When, during the World War, France rebuffed his attempts to learn about Central Powers cryptography and then failed to send back solutions of the Austrian intercepts that Italy was giving her, Sacco, who had charge of the intercept service, began to attack the messages himself. Though he knew no German, he chipped away so energetically and acutely that he soon managed to hack out fragments of plaintext. These proved valuable enough for him to be placed in charge of a cryptanalytic office attached to the Supreme Command’s intelligence service. Called the “Reparto crittografico” (“cryptographic unit”), it was staffed at first with two engineers from Irredentist areas of Austria—Tullio Cristofolini of Trent and Mario Franzotti of Gorizia—and with a distinguished linguist, Professor Remo Fedi. It employed several score of people by the end of the war.

  The cryptanalysts achieved their first complete solution of Austro-Hungarian radiograms during the Battle of Gorizia in August of 1917. What systems were then in use are unspecified. But up to that time the Austrians had not displayed any singular excellence in their cryptography. Among the systems in which they had reposed their trust and their lives was a Vigenère with alphabets normal except for the addition of ä, ö, and ü—a circumstance that perhaps explains Ronge’s vaunting of his purchase of the closely similar cifrario tascabile. There were also what the Italians called the AK and the SH, in which 50-odd ciphertext digrams represented a plaintext letter, number, or syllable. The AK was sent in its original two-letter groups, whereas the SH was divided into five-letter groups. Not till November, 1917, did the Austrians convert to codes, when they placed into service what the Italians called the cw and the Carnia codes, both of 1,000 groups and for use only within a single army. The Reparto crittografico solved them both.

  It also solved a similar code on the basis of a single message in the c
rucial days just before the Battle of the Piave. As part of the preparations for their summer push, the Austrians had placed a two-part code of 1,000 groups into service on June 15, 1918. At first they used it correctly, but soon repetitions appeared that indicated letter-by-letter encoding, with groups exceeding the frequency of 4 or 5 per cent that would be the normal maximum for word-groups in such a code. On June 20, Italy intercepted two messages with virtually the same unusual ending:

  492 073 065 834 729 589 255 073 255 834 729 264

  The pattern of repetitions suggested the plaintext radiostation, with the two partial repeats 073 … 834 729 representing the repeated a-io and the two 255s standing for the repeated t. It checked out, and thus this one lazy Austrian code clerk, who found it easier to encode letter by letter than to hunt up the codegroups for radio and station, had enabled the Italians to read a goodly portion of his comrades’ code communications.

  Italy’s growing cryptanalytic experience enabled it to solve increasingly difficult problems, such as the superenciphered Austrian diplomatic code (for which Sacco’s group had the aid of cleartext messages). The considerably larger naval cryptanalytic staff solved the Austrians’ superenciphered naval system. And gradually it dawned on the Italians that if they could read Austrian ciphers, perhaps the Austrians could read theirs. As early as January, 1917, an attempt was made to replace the old systems. It foundered on the complaint that the new methods required too much time for encipherment. Later, improvements were made to the cifrario rosso, but these were quickly nullified when a major army unit transmitted the new key variables in the old system. In June, the cifrario tascabile was replaced by a small codebook, and after the bloody Italian defeat at Caporetto, there was a wholesale change of army systems, to enciphered code. At about the same time, Cartier journeyed to Italy, visiting the intercept posts and talking with Sacco. The Allied military mission that bolstered Italy at the end of 1917 included some cryptologic personnel. All of this noticeably tightened Italian cryptologic practice.