THE CODEBREAKERS

by DAVID KAHN

  Forty minutes after midnight, the American intercept post at Souilly picked up one of the first messages in the new system. Station x2 was sending it to Station ÄN:

  00:25 CHI-13 845 422 373 792 240 245 068 652 781 245 659 659 504

  At 12:52 ÄN replied: CHI-13 OS RGV KZD. Five minutes later x2 sent a second message to ÄN:

  00:25 CHI-14 UYC REM KUL RHI KWZ RLF RNQ KRD RVJ UOB KUU UQX UFQ RQK

  When these appeared on the desk of Berthold, head of code cryptanalysis, he guessed at once what had happened: x2 sends a 13-group cipher message (CHI-13) in a new system, ÄN responds with OS, a well-known service abbreviation for Ohne Sinn (“message unintelligible”), and a reference to CHI-13, followed by two groups from the old KRU code. Whereupon x2 sends a second message, this time in KRU but with the original time group (00:25). The old KRU had been partially solved, and Berthold knew that the RGV of the short ÄN message meant “old.” He did not know the meaning of KZD, but it seemed likely in view of what happened that it meant “Send in code,” making the whole phrase “Send in old code.” Could the Germans have been so stupid as to compromise their new code within an hour after putting it into service by sending the same message in both the old and the new systems?

  Berthold’s blue eyes fairly snapped and the few pale wisps of hair that lay against his bald pate almost stood up with excitement as he decoded the second x2 message with his reconstructed KRU. It read:

  The KWZ and UOB appeared to be nulls, used—almost certainly in violation of regulations—as word dividers, and REM probably meant Kommandant. When Berthold checked this against the second message, he saw at once that it had the same plaintext. The repetitions of the plaintext i’s and t’s, which had been masked by the homophones and the lexicon of the KRU code, appeared clearly in the trinumeral message as the repeated 245s and 659s. With these four points as anchors, Berthold could set up the following equivalencies:

  A staff airplane sped his result to the British cryptanalytic bureau, and Berthold telegraphed it in a special codebreakers’ code to the French. It was a Rosetta Stone for what turned out to be the new Schlüsselheft. The three bureaus cooperated closely, but it was largely due to Painvin’s genius that within two days they had neutralized the Geheimklappe superencipherment and dismembered much of the lexicon. By March 21, when the expected German blow fell, Allied cryptanalysts were reading Schlüsselheft messages better than the German code clerks themselves. Theoretically no important information was supposed to be carried in it, because it was intended only for low-level, front-line communications. But theory succumbed at times of great activity, when the information was most desirable, and the trinumeral messages were laden with valuable nuggets. “The sending of this one message must certainly have cost the lives of thousands of Germans,” Moorman said, “and conceivably it changed the result of one of the greatest efforts made by the German armies.”

  As G.2 A.6 gained experience, it gained speed. Perhaps the most dramatic demonstration came at 9:05 p.m. April 28, when the Germans ordered an attack for 1 a.m. The message was intercepted, telegraphed to headquarters, cryptanalyzed, and employed to warn American troops half an hour before the Hun assault. Despite this rather sensational demonstration, the higher-ups seemed dissatisfied. They felt, Moorman protested, “that we were doing a lot of unnecessary work. What they wanted us to do was pick out the important messages, decode them, and let the rest go. They understood that the greater part of these messages were valueless and so thought what was the use of bothering with them. It was a matter of considerable difficulty to make them see that we had to work them out and that the Germans did not tag their important messages before sending them.”

  During the summer of 1918 G.2 A.6 received considerable help from a source that was seeking to hamper it. Code discipline had grown lax among the signal troops of the German 5th Army, which faced the American forces, and one Lieutenant Jaeger was detailed to stiffen it. He knew what should be done and issued numerous orders to do it. Unfortunately, he overlooked the circumstance that the German codebooks did not include his name, which therefore had to be spelled out letter by letter every time he affixed it to an order. This was frequently. Its peculiar formation—the repetition of the high-frequency e, for example—permitted G.2 A.6 to identify it readily, and this in turn led to important clues concerning the superenciphering Geheimklappe and, in one case, to the identification of 40 groups in a new Schlüsselheft. Perhaps it was Jaeger who, before his assignment to the 5th Army, coined one of the unforgettable slogans of communication security—which was, fittingly enough, read by G.2 A.6: Weh dem der leugt und Klartext funkt (“Woe to him who lies and radios in clear”). Jaeger was beloved by his adversaries because he kept them up to date with code changes, and it was with genuine regret that they saw his name disappear from the German traffic.

  The older, battle-wise cryptanalytic bureaus of France and England developed increasing respect for their younger protégé as it gradually proved itself, and the tendency was nurtured by the effective liaison work of Childs. A 25-year-old Virginian, he displayed enough diplomatic ability to later become American ambassador to Saudi Arabia, Yemen, and Ethiopia, and sufficient command of French to later write a book in it (on Nicolas Restif de la Bretonne, a bawdy French novelist on whom, along with Casanova, Childs became an international authority). On visits in the spring and summer of 1918, Childs built up friendly relations with Hay, Brooke-Hunt, Hitchings, and Cartier; he enjoyed a particular rapport with Painvin, under whom he studied for a week.

  Childs headed the small group that concentrated on German ciphers, a post he had obtained because Moorman mistakenly thought he was an amateur cryptologist from New York with the same name. Moorman made him liaison officer because of his impressive showing with the American trench code superencipherment—though up to that time Childs, despite diligent application of the principles he had been taught, had not solved a single German message. His ignorance was so abysmal that in London Brooke-Hunt lost all interest in him as soon as he discovered it. As a sop he told Childs about the German Für GOD cipher and furnished him with the keys. Childs soon discovered that the apparently incoherent keys were actually monoalphabetically enciphered versions of words like INSTRUMENT-ENMACHER and GOLDARBEITER. It became his personal turning point.

  On August 5, an American intercept station picked up a 456-letter message addressed to the German Foreign Office from General Kress von Kressenstein. He had recently been shifted from Syria to Tiflis to prevent the rich oil fields of the Caucasus from falling into Turkish hands. This question of friction between allies acutely interested Britain, for it affected not only her Mesopotamian campaign but also the whole future of Persia and the gateways to India.

  The intercept began: PZÄVE PNÄJY GJCGB PZAV PFAVG BPFHG YZAN RPBBP GOWIB PCBPR OOBP XBEGH ÄVBRW …. Childs took his frequency count, found to his gratification that the immense frequency of ciphertext B could only mean that it represented plaintext e in a monoalphabetic substitution, and within an hour had broken the message. Von Kressenstein was reporting the unconfirmed capture of Baku, heart of the oil basin, by Turkey. The Turkish leader, Enver Pasha, had assured him that he was moving into Baku only to improve his sanitary arrangements, and von Kressenstein was reciprocating this obvious trust. “To make the Turkish advance more difficult,” he stated, “I have hampered every shipment of munitions from Batum via Tiflis up to the present time.” Childs’ solution was important enough to be included verbatim in a printed G-2 survey of the Caucasus and Central Asia. It demonstrated to the Allies that the Turco-German split had gone deep enough for one to deprive the other of the very essentials of combat in a struggle for national existence.

  If Childs wondered in his elation why so important a message was being clothed in so flimsy a cipher, he discovered the answer a few days later when he read a message from Berlin in another and far more secure system. “The cipher method prepared by General von Kress was solved here at once. Its further use is forbidden.
”

  G.2 A.6 distributes its solutions of ADFGX cryptograms

  Extremely alert, Childs pounced on the freaks of German encipherment and twisted them to American advantage. The Germans, for example, were still using a double transposition that they called the ALACHI as one of their systems of secret communication with their forces in Russian Georgia and the Near East. On July 24, two ALACHI messages were intercepted, one of 226 letters from Berlin to a small post at Tiflis, the other of 152 letters from Tiflis to Constantinople. The Berlin message resisted all attempts at solution, but Childs discovered that the encipherer at Tiflis had omitted the second transposition. He solved it as a simple columnar with a 22-number key. When he applied this key twice to the other message, he read it with ease. It proved to be a message signed by General Erich Ludendorff, the German chief of staff.

  In another case, Childs noticed that the second part of a message from Berlin to Constantinople on November 1 was repeated the next day with the addition of only two letters. Using the slight difference as a fulcrum, Childs levered back and forth from one message to another like a bridge player working a ruff and within an hour and a half on the afternoon of November 2 had solved them both. At this time and on this front, German keys remained in effect for three days, and when he used them to read a long message of November 3 in 13 parts, he detonated a small bomb of excitement at G.H.Q.

  “By reason of its length,” Childs wrote later, “I was persuaded before the message had even been reduced to German plaintext, a long and tedious task, that it was likely to contain information of more than ordinary interest…. Every available German translator of the Section was pressed into service, while I superintended the conversion of the ciphertext to German, of which language I was almost entirely ignorant save for that instinctive feel for the mechanics of it which any cryptographer acquires from such intimate daily contact with it as I had had.”

  It turned out to be a highly revealing review of the situation in the Balkans as seen by the German commander there, Field Marshal August von Mackensen. The dynamite was in the twelfth part, where Mackensen proposed “that the army of occupation be withdrawn from Rumania at once.” The conqueror’s jackboots there were enforcing a particularly harsh peace treaty, and it was not beyond conjecture that the Rumanians would rise in fury as soon as the military restraint was removed and fall upon the Germans from behind. This possibility suddenly brightened for the Allies with the information extracted from the Mackensen telegram. Accordingly, as soon as it was deciphered and translated, Berthold grasped Childs by the sleeve and rushed with him and the telegram to the office of the assistant chief of staff. When the colonel there read it, he caught Berthold’s excitement, and he dashed out of the office carrying the telegram. He returned to say that its contents had been communicated to the Supreme War Council. A few days later, when Mackensen evacuated Bucharest to the hooting of the crowd, the Rumanian government denounced the peace treaty and declared war anew upon Germany.

  The Mackensen message was in the ADFGVX system—probably the most famous field cipher in all cryptology. It was so named because only those six letters appeared in the cryptograms,* though just five were used (no V) when the system sprang into use on March 5, 1918.

  The war in the West had by then become a stalemate of exhaustion. The young recruits who the Kaiser had promised in the glorious summer of 1914 would be “home before the leaves fall” had become veterans hardened by almost four years of battle—those few who survived. The flower of England’s youth had perished; in France, a generation had climbed out of the trenches and vanished forever.

  During the winter, Germany had come to realize that she would have to win in the spring if she were to win at all. The U-boat had failed to starve England into submission, and the United States had entered the war against her. But the collapse of Russia had freed dozens of German divisions for service on the Western Front and, for the first time, Germany held a numerical preponderance there. This, however, was only until America could transport her strong young forces across the Atlantic. It was to be now or never, and the imperial government lashed its weary troops and hungry civilians for the supreme effort that was to bring final victory.

  It was no less clear to the Allies that Germany planned to launch a climactic offensive in the spring. There were many signs—the new cipher itself was one. The question was: Where and when would the actual blow fall? The German high command, recognizing the incalculable military value of surprise, shrouded its plans in the tightest secrecy. Artillery was brought up in concealment; feints were flung out here and there along the entire front to keep the Allies off balance; the ADFGVX cipher, which had reportedly been chosen from among many candidates by a conference of German cipher specialists, constituted an element in this overall security, as did the new Schlüsselheft. The Allies bent every effort and tapped every source of information to find out the time and place of the real assault. But one of their most flowing founts—cryptanalysis—appeared to have dried up.

  When the first ADFGX messages were brought to Painvin, the best crypt-analyst in the Bureau du Chiffre, he stared at them, ran a hand through his thick black hair with an air of perplexity, and then set to work. The presence of only five letters immediately suggested a checkerboard. Without much hope, he tried the messages as simple monalphabetics; the tests were, as he had expected, negative. He discarded a polyalphabetic checkerboard as too cumbersome, and was left with the hypothesis that the checkerboard substitution had been subjected to a transposition. On this basis he began to work.

  Nothing happened. The traffic was too light for him even to determine by frequency counts whether the checkerboard key changed each day, and without this basic information he did not dare to amalgamate the cryptograms of successive days for a concerted assault. Cartier looked on over his shoulder as he braided and unbraided the letters and mused sadly, “Poor Painvin. This time I don’t think you’ll get it.” Painvin, goaded, worked harder than before. Meanwhile, Berthold achieved his Schlüsselheft entry and Painvin, shifting temporarily to that more fruitful field, completed it. But the enciphered code, used only for trench communications, provided no strategic insights. These would come, if they were to come at all, through solution of the ADFGX, which direction-finding showed was carrying messages between the higher German headquarters, chiefly those of divisions and army corps. Painvin strained even harder.

  At 4:30 a.m. March 21, 6,000 guns suddenly fired upon the Allied line at the Somme in the most furious artillery cannonade of the war. Five hours later, 62 German divisions rolled forward on a 40-mile front. The surprise was complete and its success overwhelming. French and British troops reeled back day after day in stunned confusion. The head of intelligence at French G.H.Q. came into the cryptologic bureau three days later and told Major E.-A. Soudart, the replacement for Givierge, who had gone to the front, and his assistant Marcel Guitard: “By virtue of my job I am the best informed man in France, and at this moment I no longer know where the Germans are. If we’re captured in an hour, it wouldn’t surprise me.” Within a week the Germans had punched a hole 38 miles deep in the Allied lines, and it was not until the British and French troops fell back to Amiens that they collected themselves and halted the advance.

  The furious advance was reflected in a dramatic upsurge in radio traffic. The first result was disappointment. Painvin’s frequency counts showed that the checkerboard key did change daily; presumably the transposition key did also. Solution would therefore require a goodly quantity of text from a single day, but until April 1 the interceptions were too meager. On that day, the French picked up 18 ADFGX messages totaling 512 five-letter groups. Two of them had been sent in three parts, none of the same length: the Germans had had their fingers burned by multiple anagramming early in the war and had learned their lesson.

  Studying them, Painvin noticed on April 4 that the first parts of the two messages had identical bits and pieces of text larded in the same order in the cryptograms. This oddity could
most likely have resulted from both cryptograms having identical beginnings transposed according to the same key; the identical fragments of text would then represent the identical tops of the columns of the transposition tableau. Sectioning the cryptograms so that each identical fragment started a new segment would yield the columns of the tableau, in the order of their transcription. Painvin did this to CHI-110, the 110-letter first part of a message from VI to B8, and to CHI-104, the 104-letter first part of a message sent 13 minutes later, also from VI but this time to BF:

  The problem now was to discover the transposition key, or, to put it another way, to reconstitute the block. A beginning could be made on the principle that the long columns stood at the left. Painvin saw that in both cryptograms columns 3, 6, 14, and 18—meaning the columns headed by these keynumbers—were longer than the other columns. He moved them to the extreme left of the block. Columns 4, 7, 9, 11, 16, and 19 were short in CHI-104 and long in CHI-110; consequently they clustered in a zone to the right of the first four but to the left of the remaining ten columns. The remaining ten were short in both messages and thus pushed to the right. These three zones marked a first approximation to the key.

  Unable to wrest any more information from the first parts of the messages, Painvin turned to their third parts in the hope of finding a common ending. Common repetitions showed him that they had indeed a common ending, and this enabled him to properly segmentize them into columns as he had the first parts. He partially divided up each of the three zones within themselves on the same basis of long and short columns. He thus made a second approximation to the transposition key. This showed conclusively that columns 5 and 8 huddled next to one another in the middle of the tableau and that 12 and 20 stood at the extreme right—though in neither case did Painvin yet know whether their order was 5-8 or 8-5, or 12-20 or 20-12.