Code Warriors Read online

  This, Phillips was told, was the Russian problem, and it was where he would be working from now on, and he was not to mention it even to anyone else at Arlington Hall. Phillips, Grabeel, and Lewis would likely have been astonished had they been told they would spend much of the next four decades of their lives on the problem.

  —

  The decision in 1943 to begin studying Russian traffic was odd on several counts, not least that Arlington Hall was already overwhelmed with the volume of enemy coded messages it was trying to read following America’s entry into the war.

  Even with the influx of thousands of new hires, Arlington Hall was struggling to keep up with an avalanche of operationally urgent traffic for which an hour’s delay could mean the difference between a successful battlefield coup de main or an opportunity forever lost. The solution of the Japanese army codes meant processing upwards of three hundred thousand messages a month. Reflecting the importance of that priority, a reorganization of the production branch that summer reallocated cryptanalytic work between two sections. The larger, B-II, now focused exclusively on Japanese army messages. The other, B-III, now called the General Cryptanalytic Branch, got literally everything else—all of Japan’s non-army traffic (diplomatic, military attaché); all of Nazi Germany’s traffic, military and diplomatic; and a whole kitchen sink of diplomatic and commercial codes and ciphers of some thirty other countries, enemies and neutrals alike, among them Sweden, Finland, Turkey, Bulgaria, Spain, Portugal, Vichy France, China, Mexico, Chile, Brazil, and the Vatican.16

  William F. Friedman, who for a decade after the end of World War I had been the sole cryptologist employed by the Army Signal Corps, and who then in the 1930s had patiently and painstakingly trained a cadre of young mathematicians to resurrect an Army codebreaking bureau, was one who thought there was no choice now but simply to abandon work on all diplomatic codes, much less take on any new problems, and focus exclusively on enemy military traffic. Friedman was the opposite of a defeatist. Although he had only a limited mathematical background himself—he had taken a single freshman course in mathematics as a student at Michigan Agricultural College early in the century—he had early on discovered an aptitude for codes, and a conviction that the main requirement for successfully breaking one was the confidence that it was breakable. He was the first to work out a solution to the new generation of cipher machines that used rotating wired wheels that advanced to a new position as each letter was typed in on a keyboard, generating a completely different substitution alphabet for each successive letter of the message.*3 He astonished the inventor of the Hebern electric cipher machine in 1923 by breaking ten enciphered test messages in six weeks of work, four of which he spent developing the theory for how to do it. Friedman’s precise habits and systematic approach were abundantly on display in the classic foundation textbooks he prepared in the 1920s and 1930s (Elements of Cryptanalysis, expanded into the four-volume Military Cryptanalysis), which would train several generations of cryptanalysts, bringing logic and flair to what previously had been unsystematized chaos, and in his always fastidious personal appearance with his trademark Hollywood-idol dapper mustache, perfect bow ties, and two-toned shoes. Characteristically, he played an intensely disciplined game of golf and prided himself on being an expert ballroom dancer. William Friedman, said a colleague who worked with him closely in later years, simply was never “scared of the magnitude of the problems facing him.”17

  By 1943 the organization had long outgrown Friedman’s talents as an administrator. After a hospitalization for a nervous collapse in January 1941, he was eased out of responsibilities and, at age fifty, had become something of an elder statesman and adviser. But he remained a legendary and revered figure among the rank-and-file cryptologists of the service, and the men who were now in charge of its most important technical branches were all his protégés. The three young math teachers Friedman had hired in 1930 in the first expansion of the service, and to whose training and mentoring he personally devoted the next decade, now ran virtually the entire technical side. Frank Rowlett, whose outward demeanor as a soft-spoken, courtly Virginia gentleman concealed an often prickly and acerbic contempt for incompetence, was head of B-III; Solomon Kullback, a perpetually rumpled verbal bulldozer of a New Yorker, was in charge of B-II; and Kullback’s fellow New Yorker Abraham Sinkov—he and Kullback had been classmates at City College and were both teaching in the New York City public schools when Sinkov spotted a notice for an examination for a government job as a mathematician—ran the cryptanalysis center the Army had set up in Brisbane, Australia, to bring additional manpower to bear on the Japanese army problem.

  In a detailed proposal to the agency’s chief on July 29, 1943, Friedman argued that given the duplication of effort between the Americans and the British in intercepting and processing diplomatic traffic, it made sense just to let the British have it all so that Arlington Hall could focus, without further distractions, on the much more vital Japanese army material.18

  In any case, the intelligence value of much of the diplomatic codebreaking effort was minor at best. The U.S. Army codebreakers had originally taken on the job, in fact, mainly because they had nothing else to do. Eager to give his cryptanalytic trainees some real-world experience, Friedman had found precious little else on the airwaves for them to work on during the pre–Pearl Harbor years. German and Japanese military traffic, much of it short-range transmissions sent using low-power radios, was almost impossible to intercept from monitoring stations in the United States. Diplomatic signals, by contrast, went over transoceanic commercial networks operating high-power shortwave transmitters. The decision to copy and decipher the diplomatic traffic of dozens of neutral countries had always had more to do with its ready availability than with its intelligence significance.

  There was also the fact that as sensitive as were the issues raised by spying on neutral nations, spying on an ally and now comrade in arms like the Soviet Union was something else entirely. Winston Churchill, for all his long-standing enmity toward and distrust of the Soviets, had promptly ordered a halt to the monitoring of Soviet communications by British intelligence upon Stalin’s joining the fight against Hitler in June 1941. Although no similarly explicit order seems to have been given to the U.S. Army and Navy codebreakers following the United States’ entry into the war in December 1941, both had shelved their prewar work on Russian systems given the obviously much more pressing priorities of fighting a global war across two oceans.19

  And finally there was the profoundly discouraging consideration that that earlier work on Russian codes had made precious little progress. Soviet diplomatic systems, in fact, were widely viewed as utterly unbreakable. In the 1920s, allowing discretion to be trumped by political pressures, the British government on several occasions had released the verbatim texts of intercepted and decoded Soviet cables exposing anti-British espionage, propaganda, and subversion by Soviet officials operating under diplomatic cover—and also alerting the Soviets to the weaknesses of conventional code systems. The final break in British-Soviet relations in 1927 was accompanied by a wholesale change in Soviet codes; Soviet trade missions and diplomats subsequently began using unbreakable one-time pads to encipher nearly all of their cable messages.20

  As cogent and logical as Friedman’s arguments were, they were no match for formidable political and bureaucratic imperatives whose force did not depend strictly on logic—a consideration that someone with Friedman’s orderly mind had a hard time appreciating. As early as August 1942, Captain Stevens, the British liaison officer at Arlington Hall, picked up hints that the Americans were eager to take another crack at the Russian problem, allies or no allies, and other priorities notwithstanding. Stevens, who was extremely diligent at keeping his ear to the ground, reported back to London that the Americans had never stopped collecting and filing Russian diplomatic traffic for possible future study. “Sooner or later they will inevitably try to break this,” he wrote, “since they do not trust the Russians furt
her than they could throw a steam-roller.”21

  Worries that Stalin might be secretly negotiating with the Japanese were part of that mistrust.22 But the decision by the Army in 1943 not only to retain its far-flung net to capture worldwide diplomatic communications, but to extend it to encompass the Soviet Union, had little to do with any articulable concerns about the Soviets, or even a mistrust of them in general. In the aftermath of Japan’s surprise attack on Pearl Harbor, the War Department had ordered a sweeping review of how the Army handled signals intelligence, and its major conclusion was that the way to make sure nothing ever again fell through the cracks was to miss nothing in the first place. Before the war, the Army had tended to regard signals intelligence as a technical and arcane subspecialty. It did not even come under military intelligence but had been shunted aside under the direction of the radio experts of the Signal Corps. The Navy’s codebreakers similarly came under the Naval Communications department. Ad hoc arrangements for passing on to the State Department and White House any items of importance gleaned from decoded messages had been made by the Army and Navy cryptanalytic units on their own initiative, but neither had any defined intelligence responsibilities beyond the narrow confines of service requirements. Even within the military services they had struggled for recognition and resources. Many regular officers viewed the entire business of codebreaking as a colossal waste of time. Captain Thomas H. Dyer, one of the Navy’s few trained prewar cryptanalysts, would recall with undiminished bitterness in 1945 the “general attitude” he encountered in those years, which “ranged from one of apathy to one of ridicule or open hostility….The vast majority of those whose support and cooperation were essential viewed it as the visionary project of impractical dreamers.”23

  Alfred McCormack, a formidable New York lawyer who had come to Washington right after the Pearl Harbor attack to ask his former law partner John J. McCloy, now assistant secretary of war, for “the toughest assignment he had,” was given the job of coming up with a new blueprint for signals intelligence that reflected the realities of America’s sudden thrust onto the world stage. McCormack swiftly concluded that the mission had been construed far too parochially. In an age of global conflict and global responsibilities, it was not enough just to monitor the radioed orders of enemy commanders on the battlefield. Simply, the United States “must know as much as possible about the objectives, the psychology and the methods of our enemies and potential enemies (and of our Allies as well) in order to make the right decisions.” Every country was a legitimate target, friend and foe alike; intelligence on everything from its industry to agriculture, from its politics to internal social forces, might provide the one crucial detail that would make the difference between triumph and catastrophe for American policy.24

  The wartime influx of men and money was the chance to make such an all-encompassing intelligence-gathering system a reality. It didn’t hurt that in fulfilling this greatly enhanced signals intelligence mission the War Department would also greatly enhance its prestige in the corridors of Washington political power, and its influence in postwar planning. The man brought in on McCormack’s recommendation to implement the vast expansion of the mission and scope of the Signal Intelligence Service, Colonel Carter W. Clarke—McCormack then became his deputy for the rest of the war—was in early 1942 already looking to the end of the war, and beyond, in enthusiastically endorsing McCormack’s vision. Clarke wrote in May 1942:

  Our primary task is to paint for our superiors as completely a realistic picture as possible of the activities “behind the arras” of all those associated with and against us with the end purpose of enabling an American peace delegation to confront problems of the peace table with the fullest intimate knowledge it is possible to secure of the purposes and attitudes, covert and overt, of those who will sit opposite them.25

  In a nation that would see eleven million men mobilized, a seventy-acre factory arise almost overnight on a patch of scrubland west of Detroit to begin disgorging a finished B-24 bomber every hour, fleets of cargo ships by the thousands constructed in as little as two days apiece start to finish, and the elemental forces of the universe transformed on a remote New Mexico mountaintop into a weapon of unimaginable destructive power, the idea of monitoring, intercepting, and deciphering every coded message transmitted anywhere in the world seemed neither far-fetched nor unworthy of the attempt. It was, Frank Rowlett later said, a foregone conclusion in the post–Pearl Harbor climate that the interception and deciphering of diplomatic signals would continue, that Friedman’s prudent objections would be gently brushed aside, and that Russia, ally or not, would be added to the ever-expanding list of new targets.26

  “Get everything” would henceforth be a virtual article of faith for the U.S. signals intelligence community. “Whether or not it actually can be done in practice,” insisted Rear Admiral Joseph R. Redman, the head of U.S. Naval Communications, in a memo he wrote at the end of the war, it was the fundamental mission and responsibility of the U.S. communications intelligence organizations to come “as closely as is humanly possible” to reading “every enemy and clandestine transmission.”27

  —

  There were some new practical reasons for encouragement, too, in tackling “the Russian problem.” Finland had a small but well-regarded cryptanalytic bureau, and messages from Japan’s military attachés in Helsinki and Berlin to Tokyo, deciphered and read by Arlington Hall in early 1943, revealed that the Finns had made some progress on the “unbreakable” Soviet one-time-pad diplomatic ciphers and were sharing their results with their Japanese counterparts; a series of lengthy cables offered a wealth of basic technical details about the Russian systems and how they worked.

  A considerable amount of raw Russian traffic was also now becoming available to the American codebreakers. In January 1940 the Army’s adjutant general had sent a letter to the president of RCA, David Sarnoff, asking if a Lieutenant Earle F. Cook might be assigned to the company for six months “for a course of study.” As Cook later explained, “All of this nonsense was a cover. Looking over the traffic was what I was there for.” RCA, one of the major carriers of commercial cablegrams, supplied Cook a private room and photographic equipment in its Washington offices on Connecticut Avenue; each morning he would arrive and make copies of all of the messages that had been handed in the previous day for transmission on the company’s international circuits. (Frank Rowlett would recall seeing Cook’s thumbs in many of the photos.) The clandestine arrangement—almost certainly illegal—set a precedent that would come back to haunt NSA three decades later. But no one at the time seemed overly concerned, and in any case America’s entry into World War II in December 1941 made the question of legality moot for the duration when official wartime censorship began, requiring all cables to be turned over to the government for inspection.28

  The Army was able to intercept additional Russian traffic using its monitoring stations at Fort Sam Houston, in Texas, and Fort Hunt, just south of Washington, D.C., to copy telegrams transmitted, in Morse code, over the international radio networks of other commercial operators. In June 1944, the U.S. Army negotiated an agreement with the Soviet government to establish a direct radio teleprinter link from the Pentagon to Moscow via an American radio station in Algiers; the two governments shared use of the channel and operated their own teleprinter terminals at each end. The major purpose was to improve on the unreliable and interference-prone over-the-North-Pole radio link the cable companies operated between the capitals. What the Russians did not know was that a teleprinter installed in A Building at Arlington Hall automatically copied everything passing over the circuit. For several years it would prove the most important source of enciphered Russian traffic available to the American codebreakers.29

  The Navy’s codebreakers at Op-20-G meanwhile found that they could pick up some Russian naval signals from the Far East at their monitoring stations on the West Coast. The thirty-two-year-old Navy lieutenant (j.g.) in charge of Station S at Bainbridge Is
land, Washington, was Louis W. Tordella, a math professor who had been teaching at Loyola University when the war broke out. Tordella had tried to offer his services to the Army first, visiting the headquarters of the Fifth Army in Chicago and explaining to a bored and contemptuous major that he had a PhD in mathematics, was an amateur radio operator, and had done work on codes as a hobby. The major replied, “When we want you we’ll draft you.” The Navy proved more amenable, sending Tordella a correspondence course in cryptology; shortly after passing that he received a commission and orders to report to Op-20-G in Washington. After a few months assigned to the Enigma bombe project he was sent to Bainbridge as the second in command.30

  In July 1943, Tordella received an order from Washington to put his four best intercept operators on the job of transcribing Soviet Far East Morse code traffic. It took a month for him to get hold of a single Russian typewriter. He had to send the copied traffic back to Washington on paper forms, by airmail. Most of the coded messages proved to be relatively simple hand-cipher systems used by the Soviet army, navy, police, railroads, and Communist Party and dealt with the most mundane matters: weather broadcasts, orders to icebreakers, reports on production of dehydrated vegetables.31