Code Warriors

by Stephen Budiansky

  Based as it inescapably was on piecing together information gleaned from hundreds or thousands of messages, the plain-language effort made even more of a mockery of the fiction that ASA and CSAW were not supposed to be producing their own intelligence analyses but only publishing translations of individual messages; Gurin’s group, and its smaller counterpart at CSAW, for the most part ignored the stricture and ultimately published only 0.3 percent of the plain-language messages as separate translations, concentrating instead on producing reports devoted to a series of single topics that pulled together all available information on each. Just as Bletchley’s translators and analysts in Hut 3 had done during the war, they amassed a huge card file of names, places, and industries mentioned in every message, along with other collateral information to aid the effort. This would become the nucleus of NSA’s legendary Central Reference library on the Soviet defense industry, known as C-Ref.40

  Even this limited foray into intelligence production raised hackles, especially when analysts at CIA noticed that some of the footnotes in the reports cited messages that had not been officially published and circulated. William Friedman a few years later tried, with obviously limited success, to point out the superiority of the British method:

  GCHQ has consistently strived to furnish its consumers with comprehensive reports (compilations of all pertinent plaintext material plus relevant collateral) on subjects of interest to its consumers. The full-message translation method has been reserved for the very few items which can stand alone and tell a good story. Only in the case of special subjects such as Atomic Energy has there been any attempt to publish all available messages on a subject, and in these rare instances the translations are published in book form rather than as individual cards.

  But in Washington, even the “right to write reports” was still being contested, and there was constant pressure on the signals intelligence agencies from consumers “to issue as large a number of individual translations as possible.” The Office of Naval Intelligence went so far as to insist that the signals intelligence analysts be barred from issuing any reports on Soviet shipbuilding and limit their work in that area entirely to message translation, leaving interpretation to the real experts, namely themselves: ONI informed Gurin’s boss, Oliver Kirby, that they wanted to see “everything” so that they could take over the job. Gurin protested that it had taken his group years to build up the expertise and experience needed just to identify and interpret the content and meaning of the messages they read. Many of the telegrams in fact were nothing but lists of materials and shipments and other production statistics in standardized formats that would be meaningless to someone who had not already read thousands of other similar messages and could notice connections between them. A naval officer on Gurin’s staff, Neal Carson, was so furious about ONI’s demands that he threatened to resign his commission if they persisted, which apparently so stunned some of the Navy intelligence officers who knew and respected Carson that they backed down.41

  The plain-language operation was the first plunge into the massive data collection and sifting that would characterize NSA’s standard approach in the decades that followed, and it upended business as usual. At the pinnacle of Arlington Hall’s wartime success decoding Japanese army traffic, it had processed a little over 100,000 messages a month. When the Russian plain-language effort began in 1947, ASA and CSAW were each handling a few thousand of those messages a month. The number quickly grew to unprecedented heights: 200,000 a month in 1948, 700,000 in 1949, 1 million in 1950. The peak would be 1.3 million messages a month in early 1951.42

  Simply recording, delivering, and sorting that much traffic was a challenge on a scale that dwarfed anything previously attempted. Intercepting high-frequency signals was a complex art and science that depended on atmospheric conditions, sunspot cycles, and other factors that affected how far a radio wave would propagate by skipping off the earth’s ionosphere. Not all U.S. intercept stations were yet fully equipped with teleprinter demodulating and receiving equipment needed to automatically record the signals. Intercepting Soviet telephone conversations, which also traveled by radio on many of the same links that handled teleprinter traffic between Moscow and Central Asia and the Far East, was even more difficult, often taxing translators straining to make out indistinct recordings.43

  The teleprinter traffic arrived at Arlington Hall from field sites mostly on punched paper tape, and the job of sorting and printing out those millions of tapes far more resembled working on the floor of a boiler factory than it did any contemplative scene of cerebral Holmes-like codebreaking. In an institutional culture that accorded preeminent status to the mathematically trained cryptanalysts, and where even linguists were looked down on as mere “support personnel,” the clerical tasks associated with plain-language processing were the lowest of the low. At Arlington Hall, the job fell almost entirely to a segregated unit of African American men and women.

  Up until 1944 the only African American employees at Arlington Hall were messengers and custodians. One of those was William Coffee, who had studied English at Knoxville College in Tennessee, a historically black college founded after the Civil War, and who was working as a waiter at the Arlington Hall girls’ school in 1942 when the Army took over the site. Coffee was kept on as a junior janitor and subsequently promoted to messenger. In 1944, facing pressure reputedly from the White House to improve the representation of African Americans in Army specialties—fewer than 1 percent of black selectees had been assigned to the Signal Corps, notably—the commander of Arlington Hall told Earle Cook, then a lieutenant colonel and in charge of B Branch, the cryptanalysis section, that he needed to hire “some Negroes.” The only African American Cook knew was Coffee. “I told him I got this problem,” Cook recalled. “I got to have about a hundred and some odd people of your race, and I said, ‘You’re my personnel officer to see that I get the right ones.’ Did a marvelous job. Where the hell he got them and how he got them I knew not.”44

  In fact it was not difficult. The District of Columbia had an unusually well-educated African American population, and many of the first black recruits to ASA’s cryptanalysis branch were graduates of Washington’s historically black Howard University who were working at menial government jobs well below their educational qualifications. The first segregated, all-black unit in B Branch, which Coffee joined in January 1944 and soon became the operating head of, was given the job of working on foreign commercial codes. After the war more African Americans were hired to fill low-level positions at Arlington Hall running tabulating equipment and operating key punches.45

  After 1947 the rapidly expanding Russian plain-language processing branch became the place where nearly all new African American employees ended up. The work, as a review of the agency five years later found, was “sheer drudgery.” Located on the first floor of A Building, the unit had the job of feeding the tapes into printers and stamping documents. The pay was at the very bottom of the government scale; nearly all the positions were at the GS-2 level, $1,440 a year. The one hundred or so men and women of the section, which would be known for most of its existence by its subsequent designation, AFSA-213, sardonically referred to it as “the Plantation.” It was noisy, dirty, and mind-numbingly monotonous work, but it was at least a professional position, a step up the ladder from busboys and messengers: a photograph of the unit from around 1951 shows a dozen men of AFSA-213 striking a confident and dignified pose in sharp suits and ties and gleamingly polished shoes.46

  Jack Gurin tried to improve conditions by providing at least rudimentary training in message analysis. “Their job was to sit there and watch the machine and make sure it didn’t jam,” Gurin recalled. “If it jammed, you stopped the machine and pulled out the keys or fixed the paper. Then you started it again and waited for the next jam. That was their job. They were college graduates. I looked at them and said this is ridiculous.” Gurin taught some of the staff in AFSA-213 how to read the Baudot codes at the beginning of the tape
s so that they could act as “scanners,” sorting the tapes by their addresses and selecting which were worth printing out. Scanning was not “terribly challenging,” Gurin admitted, but it was at least “better than what they were doing before.”47

  Still, even scanning was monotonous and largely mindless production-line work, a point underscored by rigid productivity quotas that required each employee to get through three hundred tapes a day, which worked out to less than two minutes per tape. The scanners were later given a fixed list of thirteen hundred terms they were to look for to decide which tapes were to be kept, which culled the number of incoming messages by 80 percent, but they were not even told what the ultimate purpose of their work was or what the Russian words meant. “So far, no other more humane or less haphazard method of reducing the millions of bits of paper to usable and workable proportions has been developed,” the 1952 agency review concluded. Although Arlington Hall had initiated an intensive six-month Russian language course in 1948, none of the African Americans in AFSA-213 were told about that opportunity; William Jones, who had studied Latin and German at school and worked in the section from 1951 to 1955, enrolled on his own in a Russian course at the U.S. Department of Agriculture Graduate School hoping to improve his chances to move up in the agency, only learning much later that the codebreaking agency offered its own course.48

  A few African Americans were able to find more rewarding positions at Arlington Hall in those years; three black electrical engineers and technicians, Carroll Robinson, Charles Matthews, and W. C. Syphax, were part of the team that built the first Abner computer beginning in 1948. Truman’s executive order that year desegregating the armed forces put an end to any legal basis for confining African American employees to separate units, and the overall political atmosphere at ASA in those years had a distinctly liberal cast. But Arlington Hall was still located in the South and segregation was the law outside the gates, including on the public buses that many of its employees rode to work. It would not be until NSA’s move to Fort George G. Meade, Maryland, in 1957, and the abolishment of NSA-63—the successor to AFSA-213—in a major reorganization in 1956 that opportunities for African Americans to escape servitude on “the Plantation” of Russian plaintext would noticeably improve.49

  What did not change was the American codebreakers’ conviction that more manpower, more machines, more intercept sites, and more money could make up, through the sheer brute force of mass collection and industrial-scale data processing, for what they could no longer achieve through cryptanalytic finesse alone.

  *

  *Dubbed “Banburismus”—the pseudo-Latin name was an allusion to the town of Banbury, where the large printed sheets used in the method were printed—Turing’s approach involved calculating the odds that a string of repeated letters that showed up in two different messages was random coincidence or the result of the same word being enciphered with the same key in both messages. It was a pioneering application of Bayesian probability to codebreaking that would become a mainstay of modern scientific cryptanalysis; see appendix D for further explanation.

  5

  Shooting Wars

  If signals intelligence was a growth industry, neither the new CIA nor the military services wanted to miss their chance to be a part of it. In peacetime, divided authority usually results at worst in turf battles, bureaucratic inefficiency, poor planning, and duplication of effort. But in the increasingly high-risk stakes of the Cold War, it was about to have literally fatal consequences.

  On April 8, 1950, a Saturday, a U.S. Navy PB4Y-2 patrol plane took off from the American air base in Wiesbaden, West Germany, passed over Bremerhaven on the coast of the North Sea, then turned abruptly east and headed to the entrance of the Baltic Sea, flying a course over the water that kept it about twenty-five miles from the coast of East Germany and Poland.1 The plane, a large four-engine craft, was a Navy version of the B-24 Liberator bomber of World War II fame, modified with a single tail in place of the Liberator’s twin vanes and an extra-long fuselage to accommodate a flight engineer’s station, radar jamming equipment, and a crew of up to sixteen for extended patrol missions over the open ocean. The Navy called it the Privateer. The crew flying that day had nicknamed their plane the Turbulent Turtle.2 It was part of a squadron usually based at Port Lyautey in French Morocco whose ostensible mission was to act as couriers for U.S. diplomatic posts throughout Europe, Scandinavia, and Western Asia.

  The Turbulent Turtle had arrived in Wiesbaden a few days earlier for what was in fact a distinctly undiplomatic mission. Like the rest of the Privateers at Port Lyautey, the plane was filled with radio gear designed to pick up radar signals from Soviet air defense units. In addition to receivers that covered the microwave radar spectrum, with frequencies up to 3000 MHz, its equipment racks were crammed with pulse analyzers and oscilloscopes that could identify the particular radar units by the duration and repetition pattern of their transmitted signals, plus direction-finding antennas to take bearings and pin down their location.

  Airborne radio surveillance went back to World War II, when the Navy and the Army Air Forces had flown what they called “ferret” flights during the island-hopping advance against Japan in the Pacific. These were extremely dangerous missions, but the information they collected allowed strike aircraft to target Japanese radar installations before each assault, an incomparable advantage in the hard-fought battles for Truk and other Japanese redoubts. After the war, the Navy sold off most of the radio equipment from its ferret squadrons as surplus. In early 1950, scrambling to reconstitute two patrol squadrons in order to begin collecting intelligence about the then almost completely unknown Soviet air defense system, the Navy sent two chief electronic technicians to New York City to see if they could buy back any of the surplus gear. Wearing civilian clothes and carrying huge rolls of cash, the two chiefs scoured the city’s Army and Navy stores and bought up all the UHF radar receivers and associated analyzing equipment they could find. The equipment was repaired by Navy technicians and quickly installed in the planes.3

  The ferret missions against the Soviet Union were carried out in the utmost secrecy, but the urgent reason for them was no secret to anyone by the start of 1950. The previous September, Truman’s press secretary wordlessly passed out to reporters summoned to his office a mimeographed statement from the president:

  I believe the American people, to the fullest extent consistent with national security, are entitled to be informed of all developments in the field of atomic energy. That is my reason for making public the following information.

  We have evidence that within recent weeks an atomic explosion occurred in the U.S.S.R.4

  On January 31, 1950, over the unanimous objection of the Atomic Energy Commission (AEC) scientific advisory committee, Truman announced in an equally terse statement that the United States would proceed with development of the “super,” or hydrogen, bomb. Up until the moment of the decision Truman held out hope that it would still be possible to place all atomic weapons under international control—such was the faith in the UN as an instrument of international order that was as yet unexistinguished in that less cynical age.

  The most nuanced and strategically insightful reservations about going ahead with the superbomb came from the AEC’s chairman, David Lilienthal, who accurately foresaw the “false and dangerous” overreliance on nuclear weapons that would skew and often paralyze U.S. defense strategy and the country’s ability to respond with military force in situations short of all-out war. But in the end even Lilienthal acceded to the tough-minded arguments of the otherwise liberal Dean Acheson, Truman’s secretary of state (who ironically was about to become the target of McCarthy’s most vicious red-baiting smears). Acheson thought the hard facts left no choice and upbraided George Kennan for trying to argue that the United States should lead by example and turn away from an uncontrolled arms race that could only end in ever-greater heights of unimaginable destructiveness. Acheson sarcastically told State Department associates t
hat if that was Kennan’s view he ought to resign from the Foreign Service and “put on a monk’s robe” and stand on a street corner, but not push his “Quaker gospel” within the department. “How can you persuade a paranoid adversary to disarm ‘by example’?” Acheson demanded on another occasion.5

  The stunningly swift move to near-total reliance on atomic weapons as the cornerstone of America’s military strategy for fighting the next war had in fact predated the Soviets’ acquisition of the bomb. Facing cutbacks in conventional forces, flush with triumphant claims of victory through strategic airpower over Germany and Japan, and locked in a fierce struggle with the Navy for postwar roles and missions and a looming budget decision between super bombers and super carriers, the newly independent U.S. Air Force seized on strategic bombardment with atomic weapons as the central pillar of its postwar planning. Its Strategic Air Command (SAC), a virtually semiautonomous force commanded since October 1948 by the hard-charging, cigar-chomping Curtis E. LeMay, was training single-mindedly to carry out the mission LeMay referred to as the “Sunday punch.” LeMay, the youngest four-star American general since Ulysses S. Grant, had earned the nickname Iron Ass from the crews he relentlessly drove in the firebombing attacks on Japan. Now LeMay insisted that SAC’s only real purpose was to be ready to deliver at least 80 percent of the entire American stockpile of atomic bombs simultaneously in a single massive strike. “The next war will be primarily a strategic air war and the atomic attack should be laid down in a matter of hours,” he proposed one month after taking over SAC, and the Joint Chiefs agreed, endorsing his plan and giving SAC the highest budget priority.6