Code Warriors

by Stephen Budiansky

  With the press hot on the trail of the mysterious disappearance of a national hero, and a major diplomatic disaster in the offing, SIS mounted a frantic cover-up, sending a police officer racing to the hotel to rip the page out of the registration book and warn the hotel owner to keep quiet, which had the predictable opposite effect. Khrushchev carried on with his itinerary, touring the Houses of Parliament, the Royal Naval College, and the British Industries Fair, and attending a special performance at Covent Garden. A poorly educated peasant’s son who had been a coal miner and factory worker before rising in the ranks of the Communist Party, Khrushchev alternated throughout his visit between anxious efforts to avoid making a fool of himself and touchy and insecure braggadocio: at one dinner party he boasted that Soviet missiles “could easily reach your island,” and he erupted in fury at a private meeting with British Labour Party officials when several pressed him on the arrests and liquidation of social democrats and trade unionists in Eastern Europe. “If you want to help the enemies of the working class, you must find another agent to do it,” Khrushchev shouted, leading one of the Labour leaders later to describe their guest as “a simple-minded man” who might possibly be qualified to hold “a secondary position in a British trade union.” But the eight-day “B & K Tour,” as the British press dubbed it, concluded without any major embarrassments on either side. Khrushchev’s jocular inquiry of Eden at one formal dinner whether he had any “missing or lost property” was shrugged off as just some obscure Russian joke.45

  It took the head of British intelligence four days to finally inform the British prime minister the facts of the matter: that Crabb had been sent on a bungled attempt to examine the Ordzhonikidze’s propeller and rudder. Two weeks later the entire story spilled out, with the Soviets lodging a formal protest of “this operation aimed against those who had arrived in Britain on a friendly visit” and a furious Eden, facing questions in the House of Commons, vowing that “disciplinary steps” would be taken and stating that his explicit instructions forbidding any espionage activities during the Russians’ visit had been disregarded.

  Exactly what happened to Crabb remains unknown to this day, but a decomposed body in a diving suit was pulled from the harbor a year later, and in 2007 a retired Soviet sailor who claimed to have been a combat diver assigned to the Ordzhonikidze said he had spotted the British frogman while carrying out an underwater patrol to protect the ship, and had slit his throat as the intruder attempted to attach a device to the ship’s hull.46

  One of the steps Eden took was to order a complete review of intrusive intelligence operations, and in particular to reassess “the balance between military intelligence on the one hand, and civil intelligence and political risks on the other.” An early casualty of the review was the Royal Navy’s signals-collecting missions near Soviet naval bases. British naval officials in Washington complained to London of their “embarrassment” at not being able to “make good our part of the bargain” to supply the promised intercepts from these missions to their American colleagues and the loss of British prestige that would result “unless we resume these activities,” but the decision had been made. When it became clear that the U.S. Navy was intending to fill the void by launching its own submarine operations off Murmansk, the British Admiralty quietly dispatched Commander John Coote, who had made the Murmansk run several times, to brief the first American crew, but only on the strict understanding that he was not to tell any other British naval officers in Washington what he was up to.47

  In spite of these unaccustomed political restrictions on its activities, GCHQ was able to remain a considerable force in the U.S.-U.K. intelligence alliance. Although it would always operate under the shadow of its far larger and lavishly funded American counterpart, GCHQ underwent a similarly explosive expansion throughout the Cold War, and for much the same reasons. The organization had been down to a virtual skeleton crew in the first year after World War II, but was soon on its way to becoming the largest and most important part of the British intelligence establishment. In 1946, GCHQ freed itself from the control of SIS and its chief, Stewart Menzies, and became a quasi-independent part of the Foreign Office. The agency would grow to more than 11,500 employees by the mid-1960s, which was not only more than SIS and MI5 put together but more than the entire British diplomatic service, including the rest of the Foreign Office staff in London and every British embassy and mission around the world. In 1952, GCHQ moved its headquarters to Cheltenham, a bucolic spa town in Gloucestershire, a hundred miles northwest of London.48

  Just as at NSA, British cryptology in the Cold War was perhaps inevitably becoming part of a muscle-bound and hidebound bureaucracy, hemmed in by rules and social and political strictures that left little room for the eccentric geniuses who had achieved their idiosyncratic triumphs against the German Enigma and other enemy code systems in World War II. Hugh Alexander had tried to lure the mathematical genius Alan Turing back to GCHQ with an offer of £5,000 to consult for a year. But in January 1952, Turing naïvely reported to the local police a burglary in his home, and under questioning about inconsistencies in his account even more naïvely revealed that the principal suspect was basically a young male prostitute he had paid to have sex with. Consensual homosexuality—“gross indecency”—was still a crime in Britain at that time, and Turing was arrested and pleaded guilty. In October of that year he confided to a friend that he knew he would never be permitted to work at GCHQ again. In June 1954, the man who had conceived the most brilliant machine-based attack on an impenetrable cipher in the history of cryptanalysis, the inventor of the logical basis of the modern digital computer, died from cyanide poisoning, either by accident or suicidal intent, at age forty-one.49

  —

  Among the administrative nightmares of the explosively growing, disjointed, and highly technical top-secret organization that Canine inherited was a notable lack of skilled managers. That was a failing common to creative and technical enterprises, which always tended to attract people more at home dealing with abstract ideas than with their fellow human beings, but it was especially acute in the very abstract world of cryptanalysis. “I had a terrible time finding people that could manage,” Canine related. “We were long on technical brains at NSA and we were very short on management brains.”50 The splintering of the work into hundreds of separate problems, each isolated technically and for security reasons from one another, exacerbated the difficulties of trying to assert managerial control on an organization made up of thousands of individualistic thinkers who marched to no identifiable drum known to management science.

  With some of the same energetic naïveté he showed in instituting the polygraph as a cure-all for NSA’s security problems, Canine enthusiastically brought in high-profile business consultants including Arthur Andersen and McKinsey and Company, commissioned organizational studies, and instituted a “management development” program that produced flow-chart-crammed publications describing in organization-science jargon the wonders of applying such concepts as “unavailability-forced decision making,” “inventories of personnel potential,” and “planned programs of experiences” to develop “the large reservoir of executive material” within the agency.51

  It was all well-intentioned but in the long run had dismal consequences, creating within NSA a large class of professional managers whose only job was to be managers, and who often had no real knowledge of the technical problems the men and women they were supposedly in charge of were working on. Jack Gurin would years later write a rueful article for the agency’s in-house publication Cryptolog entitled “Let’s Not Forget Our Cryptologic Mission,” lamenting the accumulation of a large number of positions within NSA that were “only distantly related” to its core work. A pseudonymously bylined article by another NSAer (“Anne Exinterne”) more pointedly noted the considerable overabundance of managers at the GS-13 level and above who were continually shunted from one position to another: that had been another of Canine’s hobbyhorses, as he believed th
at a regular “rotation” of managers was key to developing leadership skills. Those “left standing when the music stops” in this game of management musical chairs were invariably “assigned to a staff position with vaguely defined duties, or given responsibility for a low-priority problem, or made deputy to someone who operates very effectively without a deputy.”52

  The worst consequence of this, in the view of old cryptanalytic hands like Frank Lewis, was not just that a supervisor was all too often put “in the awkward position of making decisions affecting problems completely beyond his understanding,” but that those in such a position tended to cover up their ignorance by the familiar bureaucratic expedient of sweeping problems under the rug or putting on a Potemkin village show of efficiency. “This has reached the nadir,” Lewis complained, “when section analysts are called on the carpet for seeking the advice of staff specialists, thereby making the section ‘look bad.’ ”53

  The other nightmare Canine inherited was Arlington Hall. The hastily constructed wartime A and B Buildings had always been uncomfortable, but now they were becoming literally uninhabitable, structurally unsound firetraps. By 1954, 30 percent of the workforce had been assigned to an evening shift to relieve the impossibly overcrowded conditions. Bringing the buildings up to reasonable standard, including air-conditioning, was going to cost $2.5 million, and replacing them altogether with a new building $9.5 million, but there was a mountain of red tape involving construction of military buildings in the Washington area, as well as a new policy requiring critical facilities to be located away from possible targets of Soviet atomic bombs. In the meanwhile, the best that NSA’s employees could hope for in the sultry Washington summers was the appearance of the staff member whose job it was to whirl a wet-and dry-bulb thermometer to measure the dew point; if the combination of temperature and humidity exceeded certain maximum values (95 degrees Fahrenheit at 55 percent humidity, 98 degrees at 45 percent, 100 degrees at 38 percent), everyone would be sent home early.54

  After months of desultory discussions by the military authorities of alternative locations—Kansas City; St. Louis; Tulsa, Oklahoma; Brooks Air Force Base, Texas; Cheyenne, Wyoming; and Birmingham, Alabama were among the dozens of sites proposed—the Joint Chiefs of Staff finally issued a decision in 1951 that the agency would relocate to Fort Knox, Kentucky. Six hundred thirty miles by road from Washington, the site would presumably offer safety from an atomic attack on the nation’s capital if not much else by way of attractions. Already hinting at the discontent brewing from the agency’s civilian workforce about the proposed move, a site survey team sent to Fort Knox in August 1951 more than a little defensively insisted that “the region is neither a wilderness, nor undesirable,” and that “any normal Washingtonian can be as comfortable and happy in this area as anywhere.” The survey party likewise brushed aside the little matter of racial segregation in Kentucky, noting that “this is accomplished without noticeable friction as an accepted principle of long-established social order” and “appears to be no problem for either the whites or the negroes native to the area”—though acknowledging that it might “necessitate adjustments” for Arlington Hall’s current “colored employees,” who were not used to this quaint practice on quite the scale that it existed in Kentucky. Their report also warned against releasing “partial or incomplete” information to the staff about the move, as “a single erroneous fact or misleading statement may destroy faith and confidence in the reliability of the whole effort.”55

  That was a little late, as the faith and confidence of Arlington Hall’s employees in the whole effort was already hovering near zero. Determined to undo the decision, and with a survey of his civilian staff in hand showing that most intended to quit rather than move to Fort Knox, Canine was able to get the Army chief of staff General Omar Bradley on his side. A new site-selection board was named; as usual when Canine went to work on a problem the outcome was foreordained, and the site he had wanted all along, Fort Meade, was chosen in February 1952.56

  It was not until 1957 that NSA’s new building was fully ready for occupancy. The final cost of construction was $35 million, a 75 percent cost overrun. It was everything Arlington Hall was not: fully air-conditioned, modern, spacious, with a fourteen-hundred-seat cafeteria, five thousand parking spaces, three floors of offices along a double corridor and two wings, a basement space covering 4.9 acres to house the agency’s vast array of computers and related machinery, all on NSA’s own 950-acre site with room for expansion—which was needed almost immediately, as NSA’s growth had already exceeded the capacity of the new building by the time it was finished, requiring the communications security group to stay behind at Nebraska Avenue (where it remained until 1968).

  It was also in the middle of nowhere. Located in the still-rural Maryland countryside halfway between Washington and Baltimore, the main thing Fort Meade had to offer was space; the post, which had served as a major training base in both world wars, had hundreds of wooden barracks but not much else in the way of permanent facilities. There was only limited housing available on post, requiring most of NSA’s staff to commute. The new Baltimore–Washington Parkway eased the twenty-mile drive from downtown Washington, but the other roads leading to Fort Meade were tiny, winding two-lane rural routes, and there was only limited public transit available, the only real option being a one-hour-and-twenty-three-minute slog by train from Washington’s Union Station and connecting bus from Laurel, Maryland.

  A cartoon from the NSA newsletter; the “Meademobile,” parked at Arlington Hall, provided employees information about the Fort Meade area in advance of the agency’s move there in 1957.

  To win over still-reluctant NSAers, Canine was able to have the relocation declared a “permanent change of station,” permitting the government to cover the full cost for civilian employees to move their household effects. A trailer dubbed the “Meademobile” was parked in the picnic area between A and B Buildings at Arlington Hall and staffed by cheery uniformed young women attendants who offered information about houses and apartments, schools, churches, and shopping in the Fort Meade area. In the end only 2 percent of NSA’s staff quit because of the move. Canine retired just before NSA took up quarters in its new home. The myriad details of the relocation had almost fully occupied his final two years as director.

  Canine’s successor was Air Force lieutenant general John A. Samford. He was virtually Canine’s opposite, a cerebral intelligence professional who fully understood the world of signals intelligence, “more of a pedant than a pilot, more of a philosopher than a fighter,” in the words of an internal CIA history. Samford probably wished he had something as trivial to worry about as the logistics of relocating ten thousand people and hundreds of tons of sensitive equipment and classified files, compared to the challenges that greeted him as he walked into the new director’s office on the second floor of the NSA Operations Building for the first time: above all, the codebreakers’ continuing inability to do what they were supposed to do to justify their half billion dollars a year, namely to actually break the codes of America’s chief adversary, the Soviet Union.57

  7

  Brains Versus Bugs

  NSA’s failure to read much if any Soviet encrypted traffic since 1947–48 was obviously becoming more than just a temporary setback: something fundamental had changed in the nature of the Russian cryptographic systems, and in the eyes of some scientific experts called in to assess the situation, NSA had failed to keep up with the times. The Brownell Committee, whose report back in 1952 had led to the creation of NSA, circumspectly observed that “efforts in certain important parts of the cryptanalytic field have not been crowned with success, to say the least,” and suggested that a completely new approach was called for. “The attack is timid, parsimonious, and too bound by the remembrance of past accomplishment to make the fresh and untrammeled start that is demanded,” the committee reported, citing cryptanalytic experts outside the agency they had interviewed. The Robertson panel reiterated the
point the following year: “Top priority should be accorded to the solution of high-level Soviet cryptographic systems.” And in 1955 yet another outside panel, this one headed by retired Army general Mark Clark, part of a comprehensive review of government that Eisenhower asked former president Herbert Hoover to carry out, urged that given the “vital” importance of communications intelligence as insurance against a surprise atomic attack, an immediate effort “at least equal to the Manhattan Project” was justified to regain the ability to read high-level Soviet systems.1

  Despite the millions of dollars invested in computers and special-purpose analyzers, no breakthroughs had been achieved. The very first computer program written for Atlas was designed to search the Russian one-time-pad traffic for repeated key use, and a GCHQ special-purpose digital comparator, Oedipus, joined that effort in 1954. Oedipus was basically an electronic improvement on the U.S. Navy’s war-era relay and punch card Mercury machine. It had a stored read-only memory that used thousands of semiconductor diodes to assign a weight to each of four thousand known code groups, while a magnetic drum memory held ten thousand strings of hypothetical key derived from the first three words of accumulated message traffic. A paper tape containing the first three words of a series of new messages to be tested was read in, each was stripped in turn of each of the ten thousand possible sequences of key, and the resulting code group values were looked up in the dictionary and the weight score recorded to see if any represented possible key-page reuses. This machine-aided analysis added to the discoveries of additional depths in the mid-1940s messages, but despite extensive searches no later instances of exploitable one-time-pad traffic were uncovered at all.