War of Shadows

by Gershom Gorenberg

  When Balbo died, the Duce sent General Rodolfo Graziani back to Libya to replace him. Mussolini had two great worries that summer. The first was that Britain would ask Germany for peace. The second was that Germany would conquer Britain quickly, before Italy could overrun Egypt and unite the pieces of its African empire. That fear grew in August as waves of Luftwaffe bombers pounded England and fought the RAF in preparation for landings.

  In Italy the winter before, Graziani had wanted to join Germany in the war. In Libya, his enthusiasm dried up. Visiting Rome in August, he told Foreign Minister Ciano how unready the army in Libya was. “We move toward a defeat which, in the desert, inevitably becomes a rapid and total disaster,” Graziani said. If he absolutely had to send his divisions forward, he wanted to wait for cooler weather.32

  From Berlin, Mussolini got “significant hints” that “the decisive attack against Britain” was about to start. It did not occur to him that the British would last longer than the French. Mussolini expected both “victory and peace” by the end of September, and sent orders to Graziani that he must attack—certainly by the day when “the first platoon of [German] soldiers” landed on an English beach.

  At last, the offensive was set for September 6. When the date came, Graziani asked for another month. Mussolini now thought that the war might last all the way into winter. But he’d been living in expectation for too many weeks. Graziani must attack or be replaced, he told his ministers.

  The general followed orders.

  On September 9, three months after Mussolini’s decision to go to war, the Italian Tenth Army rolled toward the border with Egypt.33

  2

  WAR OF SHADOWS

  Summer–Autumn 1940. Washington–Bletchley Park–Rome–Budapest.

  IN ONE CORNER of the Munitions Building, the US War Department’s massive warren of offices near the Lincoln Memorial, Roosevelt’s appointment of a new secretary of war brought a surge of anxiety. As secretary of state in 1929, Henry Stimson had closed down the Black Chamber, which cracked codes and read other countries’ diplomatic cables. In the summer of 1940, William Friedman and his small band of codebreakers at the army’s Signal Intelligence Service feared that Stimson would shut their operation as well, just as they were on the obsessed, manic edge of a breakthrough.

  Back when Stimson shut the Black Chamber, Friedman was left as the only code expert in the War Department. Like virtually everyone else who worked in cryptography in those days, he came to the field by accident.

  Friedman was born in 1891 in Kishinev, a half-Romanian, half-Jewish town at the western edge of the Russian Empire. When he was a year old, his family emigrated to the United States to escape the threat of anti-Semitic violence in the tsar’s realm. (Many years later, a psychiatrist treating him would note that Friedman lived with a constant, inherited fear of pogroms.) Overcoming the quieter but endemic anti-Semitism in America of his day, Friedman succeeded in getting into Cornell University. With his degree in genetics, he was hired by a millionaire who had set up an agricultural lab as a personal hobby. Then Friedman was put to work in his benefactor’s other project, looking for the code in Shakespeare’s writing that would supposedly reveal the “true” author. In the process of disproving that theory, Friedman invented mathematical principles for codebreaking. He also came to the understanding that applying rules could take you only to the edge of revelations. To crack a code, he’d write, you also had to “let your subconscious solve it.” His own intuitions may have been picked up from his father, who’d been a linguist and translator before the move to America. Solving codes—as Dilly Knox best demonstrated in Britain—can be translation taken to a higher level: puzzling out meaning from a language about which you know nothing.1

  In 1921 Friedman moved from the millionaire’s estate to the Munitions Building on Constitution Avenue in Washington to become the chief cryptographer—and the only cryptographer—of the Office of the Chief Signal Officer of the US Army. A photograph shows Friedman with a high forehead, sculpted cheekbones, and a narrow chin. His wide-set, hooded eyes are melancholy, as if sending a coded message to ignore the smile on his lips. After the Black Chamber closed, he was allowed to hire three assistants. In contrast to their British counterparts, they did not come from exclusive universities or from an aristocratic old boys’ club. US government rules required that Friedman hire only people who’d passed the civil service exam. Two of the three were young Jewish mathematicians who’d been teaching school in New York. They were old friends from City College, the academic paradise of New York’s teeming immigrant masses yearning for higher education. Abe Sinkov brought Solomon Kullback a flyer he’d spotted, saying that the civil service was looking for mathematicians. Both were frustrated as schoolteachers; they wanted to use their math. The civil service, though racially segregated, was a refuge for Jews for whom jobs in corporations and universities were nearly off-limits.2 The third man, Frank Rowlett, had been teaching math in the back country of southwest Virginia. When Rowlett got a letter offering him a job as “junior cryptanalyst,” he couldn’t find the word in the dictionary. Friedman explained afterward that it meant “codebreaker” and that he’d invented it.

  The Signal Intelligence Service, or SIS, grew very slowly. Officially, Friedman wasn’t supposed to engage in “the solution of… secret or disguised enemy messages” in peacetime. He was only allowed to train his people to do that, so they would be ready for war. In the meantime, their task was developing codes for the army and government agencies. To create one—apparently the Military Intelligence Code, to be used by military attachés—they wrote the fifty thousand words that needed to be in the codebook on one set of cards and an equal number of meaningless groups of five letters on a second set. Then they went inside a room-sized vault, tossed all the cards of one set in the air, picked them up one by one, and matched them with cards from the other set.3

  They used letters rather than numbers, apparently because attachés sent their reports to Washington via commercial firms. Radiograms came into Washington via RCA, the Radio Corporation of America. The companies had historically charged less for letters than for numbers. The catch was that with letters you couldn’t use subtraction tables—lists of random numbers to subtract from the code groups in order to disguise them. Friedman came up with another solution: tables in which the alphabet from A to Z appeared at the top, followed by line after line in which the alphabet was scrambled. If the first letter of the coded message was K, the sender would replace it with the letter that appeared beneath K in the first line of the table. For the second letter, he’d use the second line. Tables could be produced quickly, unlike a new code. So each attaché could get a different set, and they could be replaced regularly. The system was a poor man’s Enigma: slowly, by hand, each letter was encrypted in a different cipher.4

  Meanwhile, under the cover story that they were only training at cryptography, Friedman and his SIS staff worked on the new codes that Japan adopted after Herbert Yardley had written his tell-all book American Black Chamber, which had warned the Japanese more than a decade earlier that their codes were compromised. In the mid-1930s Friedman’s team succeeded in cracking the first, simple cipher machine that the Japanese used for diplomatic traffic. The Americans code-named it Red, and it provided a stream of intelligence on Japanese intentions. Toward the end of 1938, a disturbing message appeared. It authorized “travel for a ‘communications expert’ named Okamato” to the most important Japanese embassies around the world, including Washington, London, Rome, and Berlin. His job was to set up a new cipher machine. On February 19, 1939, another intercepted Japanese message announced that the new device would go into use the next day. With American-Japanese relations growing ever more tense, the vital intelligence stream dried up.

  SIS named the new machine Purple. Friedman’s team could see that it was based on the Red machine, but was far more sophisticated. At the Munitions Building, the SIS mathematicians began looking frenetically for a solution. Fri
edman, whose job had slowly shifted to administrator, received orders to get back to work on codebreaking.

  The characteristics of Red that carried over to Purple provided some clues. Both machines were used to send messages in Japanese transliterated into English characters. Red had encrypted vowels separately from consonants. Purple still encrypted six out of twenty-six letters in an easily broken method, though the six letters changed daily. If you figured out the six for a given day, it was sometimes possible to work out much more of the original, plain-language message. Occasionally, the Japanese made the cardinal error of sending the same message in Purple to the most important embassies and in Red to other diplomatic missions.

  The mystery was how the other twenty letters were being encrypted. When war broke out in Europe, the SIS got a budget increase that brought Friedman two dozen new staffers, including Genevieve Grotjan, a brilliant twenty-six-year-old mathematician who’d been unable to get a college teaching job—very likely because of another endemic prejudice, against women. In contrast to its British counterpart at Bletchley Park, the SIS paid women the same as men.

  Then Stimson became secretary of war.

  The chief signal officer talked to Roosevelt’s military aide, and received dispensation to keep the resurrected Black Chamber secret from the secretary. By mid-1940, the Purple team was working deep into every night, coming in on weekends, and borrowing staff from other projects. Friedman himself woke earlier and earlier, drinking coffee and pacing his kitchen for hours before going to the Munitions Building. There, in the humid heat of the Washington summer, the sweaty codebreakers fought the fans in their rooms to keep the papers on their desks.

  On a September afternoon, Gene Grotjan had a breakthrough. Staring through her rimless glasses at comparisons of Purple messages and plain-language texts, she realized that the twenty-letter sets each day were going through a triple scrambler, similar to the three rotors of an Enigma machine. She rushed into the next room, where the senior staff worked, to ask Rowlett to come see. “Gene has found what we’re looking for!” Rowlett shouted moments later. Cheers erupted in the office. Grotjan’s glasses enlarged the tears in her eyes. A week later, her insight allowed the codebreakers to decrypt and translate two messages.5

  It was September 20, a week before Japan, Germany, and Italy signed the Tripartite Pact, binding them to mutual defense against attack on any of them. The timing of Grotjan’s breakthrough was not coincidence: the negotiations apparently led to a surge in Japanese diplomatic traffic, providing the SIS with more raw material.

  At the time, it appeared that the three Axis countries, especially Germany, created the pact to deter America from entering the war. The deeper explanation was that on the chessboard of Hitler’s megalomania, the pact was a move meant to lead Japan to destroy the American navy while the Führer was busy with other plans.6

  Three weeks later, under the direction of Lieutenant Leo Rosen, who’d been assigned to the SIS because he’d studied electrical engineering at MIT, a signal corps team had built an ersatz Purple machine. They’d never seen the real thing, but their copy could decode messages—if you knew the daily settings. The Japanese were only using a few hundred of countless possible combinations. Friedman’s report from mid-October 1940 says that his staff had already worked out a third of the settings actually in use. The decrypted messages were appropriately code-named Magic.

  Word of the wondrous SIS device reached the army chief of staff, General George Marshall. He visited the SIS to see it working. Then he brought Henry Stimson.

  The secretary of war watched the making of Magic and liked it. He did not shut anything. As he’d later explain, “What you do in war and what you do in peace are two entirely different things.” Even if the United States was not a belligerent at the moment, he’d been appointed to his current office to prepare for the looming possibility of war. Magic allowed Marshall, Stimson, and Roosevelt to stand invisibly behind Japan’s diplomats in Rome and Berlin, Washington and London, and read over their shoulders.

  Friedman, who’d been in the army reserves for years, was called up for active duty at the rank of colonel. Two weeks later, he had a nervous breakdown and was taken to the army’s Walter Reed Hospital. Extreme nervous fatigue, the doctors concluded, which perhaps should be decoded as saying there is only so much sleepless strain the subconscious can sustain. Having succeeded in his mission, Friedman collapsed.

  MARIAN REJEWSKI WAS now Pierre Renaud, a teacher from Brittany. So his papers said. It’s likely that whoever forged them chose Brittany in the hope that no provincial Vichy policeman in the small town of Uzes would realize that “Renaud’s” accent was shaped by Polish, not by Breton, a language that was still spoken in a far corner of France. Jerzy Rozycki’s documents said he was Julien Rouget. Henryk Zygalski had become René Sargent. Gustave Bertrand of French military intelligence bought a chateau outside Uzes under his latest alias, “Monsieur Barsac.” Command Post Cadix, as it was known to a very few Frenchmen, began operating on October 1, 1940. To the Polish General Staff in exile in London, it was Eksopytura 300—Field Office 300.

  The journey that ended in Uzes had begun on June 10 at 1:40 a.m. at another chateau, the one secretly known as Command Post Bruno, at Gretz-Armainvilliers, twenty-five miles from Paris. By the time the sun rose and set that day, Italy would have joined the war. Much more important for Bertrand, German armies would be even closer to Paris. Bertrand had orders to evacuate the fifteen members of the Polish Cipher Bureau working for him on German Enigma, along with the seven Spanish Republican codebreakers who decrypted the ciphers of Italy and of the Franco regime in Spain.

  Bertrand’s orders did not come with vehicles. He commandeered a truck and a bus, loaded them with the post’s staffers and files and Polish-made Enigma machines, and headed out into the churning current of refugees on the roads southward.7 Rejewski, Rozycki, and Zygalski were fleeing a German invasion for the second time in nine months. Even more important than preserving their lives and their ability to break Enigma was keeping the knowledge that it could be broken from the Nazis.

  A hundred miles south of Paris, at La Fèrte-Saint Aubin, they set up shop—and then fled further south after the fall of Paris to Le Chatelet, then onward to Vensat, and to Agen, moving almost daily. They were at yet another temporary refuge, a wide spot in the road called Bon Encontre, when news came of the French surrender.

  The armistice required France to hand over to Germany the names of everyone engaged in signal intelligence. Bertrand managed to get his fugitive Poles and Spaniards onto three planes from Toulouse to Oran in Algeria. From there they moved on to Algiers, where they registered under false names at the Touring Club Hotel, avoided talking to locals, and remained in virtual confinement for close to three months. The Abwehr and the SD—the Nazi Party’s intelligence bureau—were spreading out in the officially unoccupied part of France administered by the Vichy government and in Vichy-controlled Algeria.

  Under the armistice, the French army had to shrink to one hundred thousand men. Within that small number, a large tangle of conspiracies and conflicted loyalties grew. Bertrand sided with a covert anti-German faction and decided to reestablish his intelligence operation on Vichy territory. His plan was approved by General Maxime Weygand, defense minister of the Vichy government, a man reasonably suspected by both the Allies and the Germans of working against them. Weygand served in the collaborationist regime, but wanted his own sources on what the Germans were up to.

  With their forged papers, the Polish and Spanish codebreakers returned from Algeria and got to work. The only radio intelligence service that the Germans allowed in Vichy was an agency that was supposed to search for Resistance transmitters. Pro-Resistance officers working there instead monitored German traffic and provided intercepted messages to Command Post Cadix. Inside occupied France, the Resistance tapped telegraph cables and sent couriers to Uzes carrying encoded Nazi telegrams. At the Uzes chateau itself, four Polish wireless operators liste
ned through the shortwave static to the Abwehr and SD radio nets. Rejewski and his colleagues broke the codes. From Cadix flowed an underground stream of warnings to the Resistance of what the secret police were planning. Via a shortwave link, the Polish General Staff in London received intelligence it could share with the British.

  But the personal link to Bletchley Park was broken. Bertrand could not come to England bearing gifts; Alan Turing could not cross the Channel to talk to the Poles.

  In the summer and fall of 1940, Britain was engaged in two fateful battles. In one, the Battle of Britain, the German air force intended to destroy the Royal Air Force and break the British will before a Nazi amphibious invasion. In the other, the Battle of the Atlantic, Britain was trying to bring the convoys of arms and supplies from America that it needed to survive, as German submarines moving in packs hunted its ships. It was winning the first battle and losing the second.

  Unknown to the Germans, the press, and the British public, the war of the mind at Bletchley Park also had two fronts: against the Enigma methods of the German air force and against German naval Enigma.

  On the first front, Bletchley Park was gaining ground, as it began to fulfill Welchman’s vision of industrialized codebreaking. The first of the new bombes was delivered in August, its key-breaking ability speeded up with Welchman’s diagonal board. The bombe’s limitation was that it depended on a crib, a very educated guess about a short bit of text likely to appear at the beginning of a message. The guess depended on knowing what was in earlier messages.8 So it was most effective against the German air force key, Red, which Hut 6 had broken and read in the past.9