Code Warriors

by Stephen Budiansky

  It was all something of a nightmare, and a massive program called TEMPEST was launched to try to develop effective shielding or masking technologies. But many of the problems defied easy solutions. Engineers tried to obscure the stray teleprinter signals by having all five relays that registered the bits of the plaintext letters actuate at once, rather than sequentially, thus producing a single spike instead of five smaller ones, but they found they could still ascertain the letters by measuring the total size of the single spike. Attempts to muffle sound coming from a code room by lining it with acoustical tile actually turned out to make the problem worse, dampening reverberations that obscured the distinctive machine noises. Side channel attacks, and the defenses devised to thwart them, would become an increasingly dominant part of the signals intelligence wars between the United States and the Soviet Union as advances in cryptographic sophistication stymied the traditional tools of the codebreakers.30

  Making an end run around cryptanalytic solutions seemed to many the only hope in any case. The report of the Baker Panel in the late 1950s pointedly balanced its pessimism over the prospects for ever solving Soviet high-level systems through business-as-usual cryptanalytic attacks with the observation that there is “no limit” to what can be obtained through the CIA’s more direct approach: “If machine plans, key usage schedules, and operational information is stolen, any system can be read.”

  It particularly urged the employment of such second-story cryptanalysis for advancing the so far fruitless efforts to crack any post–Black Friday one-time-pad Soviet traffic: “All permissible efforts should be made to obtain one-time key material.”31 An expanding program of surreptitious entries at foreign embassies around the world, aimed specifically at copying code materials and planting bugs, became CIA’s major contribution to the effort after the end of the Berlin Tunnel operation.

  Following Canine’s retirement, Rowlett returned to NSA as a special adviser to the director. He had always maintained that production of intelligence was not actually the most important aspect of cryptology: far more vital was safeguarding one’s own secrets.32 It was now beginning to look as if that was all cryptology had to offer in any case.

  —

  On February 25, 1956, Khrushchev delivered to the 20th Congress of the Soviet Communist Party an astounding indictment of the crimes of his predecessor, cataloging Stalin’s purges and “cult of personality.” When the Polish Communist Party leader Bolesław Bierut read the text of the speech, he had a heart attack and promptly dropped dead. By June, CIA had obtained a copy of the initially secret document and leaked it to the New York Times. It was, said Allen Dulles that summer, “the most damning indictment of despotism ever made by a despot.”33

  The new Soviet leader was, in the assessment of John Lewis Gaddis, “genuine—and fundamentally humane—in his determination to return Marxism to its original objective” of offering a better life for people than did capitalism. But Stalin’s ghost was not to be “so easily exorcized,” for the immediate effect of Khrushchev’s call for reform was a wave of rebellion within the Soviets’ overextended empire, posing an immediate challenge that seemed to leave little choice between crushing force or compete loss of Russian control.

  Taking Khrushchev’s speech at its word, the Polish government began freeing political prisoners, including the former Communist Party leader Władysław Gomułka, who had been deposed by Stalin. In October, without seeking Moscow’s approval, the Polish Communist Party met in Warsaw and chose Gomułka as Bierut’s successor.

  Khrushchev’s desire to break with his predecessor’s brutal methods and to give socialism “a human face” was real enough, but he had not risen through the ranks of the Stalinist regime without learning how to deal ruthlessly with those who threatened his power. He had had Beria shot after Stalin’s death, correctly viewing him as a dangerous rival, though ironically on the accusation that Beria was too willing to reach a rapprochement with the West: Beria had proposed accepting a reunified and capitalist Germany if it remained neutral and outside the NATO alliance. (Beria had also nearly dismantled the entire foreign operations of the MGB/MVD in 1953, recalling or dismissing its residents from embassies in every Western country and pulling back seventeen hundred officers from Germany alone. This was likely an effort to remove officials loyal to his rivals, but it gave his enemies a further opportunity to accuse him of being an agent of “English and German intelligence.”)34

  Furious over Poland’s independent moves, Khrushchev flew uninvited to Warsaw, threatened an invasion by the Soviet army, and relented only when Gomułka pledged to continue the military and political alliance with the Soviet Union. But that same day, October 24, protests in Hungarian cities against that country’s hard-line Communist regime became an outright armed rebellion.35

  The ongoing impenetrability of Soviet code systems left NSA to rely upon traffic analysis, direction finding, and plain-language communications to follow the chaotic and uncertain events in Hungary. An order sent in the clear to the 2nd Guards Mechanized Division of Soviet Forces in Hungary on the morning of the twenty-fourth was intercepted by an ASA monitoring post at Bad Aibling in Germany: it ordered the unit, normally stationed fifty miles south of the capital, to move without delay to Budapest and use its tanks’ cannons against the “rioters.” Other signals over the next several days pointed to heavy Soviet casualties and growing hesitation on the part of commanders. After several days of bloody street fighting in Budapest, Soviet army forces that had initially come to the aid of the government pulled back, and it seemed as if Khrushchev was even considering letting Hungary fall out of the Soviet orbit.

  On the night of November 2, direction-finding fixes on radio traffic associated with identified Soviet ground force units, however, showed a massive movement of troops within the country and fresh reinforcements pouring across the border, indicating that a final Soviet offensive to crush the uprising was imminent. On the morning of November 4, the Soviet army attacked Budapest and other cities, and the Russian generals who were engaged in negotiations with Hungarian officials to end the Soviet intervention “abandoned the pretense and arrested the conferees,” CIA reported. Prime Minister Imre Nagy issued a desperate plea for Western military assistance before he, too, was arrested. He was later shot, adding one more casualty to the twenty thousand Hungarians and fifteen hundred Soviet troops killed.36

  Amid the Hungarian uprising, Britain, France, and Israel on October 29 launched a secretly planned and ultimately illfated military operation in the Sinai, adding another crisis to rapidly escalating international tensions.

  The action was ostensibly limited to returning control of the Suez Canal, vital to Britain’s and France’s oil supplies, to the internationally owned Canal Company following Egyptian president Gamal Abdel Nasser’s surprise declaration in July that he was nationalizing the waterway. But it was more broadly intended to deal Nasser a humiliating blow that would lead to his overthrow. Smacking of unreconstructed European colonialism, the Sinai operation infuriated Eisenhower—he had not been told what America’s allies were up to—and gave Khrushchev an opportunity to bolster the Soviet position on the world stage by threatening to strike London and Paris with “rocket weapons” if their forces were not withdrawn at once. Behind the scenes, Eisenhower threatened the invaders with economic sanctions, which proved a more believable and effective message. On November 6 the British government, under unprecedented pressure from Washington, agreed to a cease-fire, and by the end of the year the Anglo-French Task Force was withdrawn, replaced by UN peacekeepers.

  In his memoirs, Eisenhower insisted that “we were in the dark” about what Britain and France intended to do, that he was unable to “fathom the reason” why Israel was mobilizing its forces, and even after Israeli troops crossed into the Sinai did not believe the invasion was part of a plot involving Britain and France. Secretary of State John Foster Dulles, Allen Dulles’s brother, told a press conference in December 1956 that the attack “came as a com
plete surprise to us.” But this was the administration’s effort to underscore its dissociation from the action and disapproval of it, rather than an honest reflection of the intelligence that the United States had at its command. Allen Dulles was furious over his brother’s remarks, later pointing out to a reporter that U.S. intelligence had accurately foreseen the attack, knew that all three countries were involved, correctly identified the intention, and had reported the day before that it was imminent. As with the Hungarian uprising, most of the intelligence came from NSA, which had followed the Israeli mobilization and the deployment of British troops to Cyprus; during the last two weeks of October, NSA also noted an upsurge in diplomatic communications between Paris and Tel Aviv and between London and Paris, leading it to conclude that France was planning “actions in conjunction with Israel against Egypt.”37

  Strikingly, as much as the event strained U.S.-British diplomatic relations, it had no discernible effect on cooperation between NSA and GCHQ at the working level, which “continued without interruption” throughout the crisis and afterward, according to an NSA internal history—underscoring both the exceptionally strong personal ties that had grown up between the two agencies over the course of a decade and a half of technical collaboration and the unique nature of the UKUSA signals intelligence agreement, which seemed to occupy its own hidden world of international relations completely apart from the normal channels through which the United States conducted its foreign policy.38

  NSA had acquitted itself reasonably well in providing signals intelligence coverage under far from favorable circumstances. One of its most important findings was a negative one: the absence of any indicators that Soviet forces were mobilizing to intervene in the Suez Crisis, reassuring Eisenhower that Khrushchev’s bellicose threats against Britain and France were a bluff.39

  But trying to put the agency on high alert to respond to the twin crises in the fall of 1956 laid bare just how utterly unprepared NSA was to shift gears in an emergency. The entire organization had been built as a vast assembly line, with the work broken down into individual, specialized tasks; there were few generalists who could jump into a new problem or take over an urgently needed task that demanded extra attention. Emphasizing the factory-like structure, the “production” branch, called PROD, was divided into four sections: GENS, which handled the “general Soviet” problem; ADVA, focused on advanced work on Russian ciphers; ACOM, responsible for Asian Communist countries; and ALLO, which revealingly stood for “All Other.”

  There was a plan on paper to place NSA on four different “COMINT Alert” levels in the event of an emergency. Alpha was the highest, for an actual war, followed by Bravo, X-ray, and Yankee. A Yankee Alert was intended to increase coverage and reporting when “planned U.S. or Allied activity may stimulate a foreign communications reaction or provoke military or paramilitary action by a foreign nation with respect to the U.S.,” and it called for issuing intelligence reports at six-hour intervals at an “intermediate” level of precedence. But even that immediately strained NSA’s capabilities beyond its limits when Canine declared a Yankee Alert in the midst of the Hungarian and Suez crises. Canine was flabbergasted to learn it took seven hours for his order even to be distributed: no one had ever worked out the procedure for what to do if an alert was actually declared. Other field units meanwhile kept filing their normal daily reports, clogging the lines and delaying delivery of the alert material, which was supposed to have precedence. The entire communications system that NSA depended on to move intercepts from the field to Washington virtually ground to a halt under the surge of reporting, an inadequacy that a postmortem of the alert said was “appallingly apparent.” The agency’s internal history of the era bluntly concluded, “As for crisis reporting, all was chaos. The cryptologic community proved incapable of marshaling its forces in a flexible fashion to deal with developing trouble spots.” All the more so if those “trouble spots” ever involved U.S. troops in another real war, where not just days or hours, but minutes, mattered.40

  —

  The larger problem was that the relentless focus on breaking the unyielding Soviet ciphers was locking the entire agency into a bind. The Baker Panel suggested that radical surgery was needed: ADVA was sucking so much effort from other departments that it was threatening the health of the whole organization and it ought to be cut out, to be turned over to a Los Alamos–type think tank devoted solely to long-term mathematical research. “The intellectual problem is much too refractory to yield to administrative pressure, and extreme emphasis on this one project hampers the NSA and belittles its many valuable contributions in other directions,” the panel concluded.41

  Past wartime successes…have established an ideal image, a standard, a set of values in the NSA which is reflected in its organization and operations, but which is not appropriate to the realities of today. What is needed now is a complete division between cryptanalytic research on [high-level Soviet ciphers] and the actual production of current intelligence.42

  William Friedman was practically in tears and threatened to go to the White House when he learned of the recommendation. It was, of course, personal: Friedman had practically invented the modern science of cryptanalysis, had built the first truly professional codebreaking organization in the U.S. government, and had himself directly trained many of the men and women who were now NSA’s top cryptanalysts. It was hard not to read Baker’s report as saying that they just were not good enough anymore. But Friedman also saw cryptanalysis as the heart and soul of signals intelligence; it was what gave the agency its intellectual energy and distinct sense of purpose, and he thought it was a fundamental mistake to remove cryptanalysts from direct contact with operational needs and sequester them in some remote, theoretical, academic elysium.43

  Yet Friedman, truth be told, was part of the problem himself: as NSA’s chief liaison to the scientific and mathematical community through the 1950s, he repeatedly maneuvered to prevent the agency’s top academic consultants—the members of its Special Cryptologic Advisory Group, later the NSA Scientific Advisory Board—from learning much about, much less contributing to, the top-level Soviet problems that had so stubbornly resisted NSA’s own efforts at solution. This group of distinguished outside mathematicians and scientists met in Washington only twice a year for two or three days, during which time they were treated to dog and pony shows with little opportunity for making substantive contributions; most of their advice ended up dealing in generalities about developing the agency’s scientific and mathematical expertise. Members repeatedly complained about being brought in “to hear a couple of lectures” but not actually working on any specific problems.44

  Even participants in a more substantive program to bring the nation’s top mathematical brainpower to bear on NSA’s unsolved problems found themselves getting the runaround. The Special Cryptologic Advisory Math Panel, or SCAMP, was begun in 1952 as an annual summer symposium held on the campus of UCLA. But the agency restricted discussions to the Confidential level, and the initial topic NSA chose for the group to examine, an abstruse area of pure mathematics known as finite projective planes, had at best a “tenuous” connection to cryptology, according to a subsequent assessment. Several years later, when panel members protested that NSA was giving them little more than “sales presentations” while denying them any essential information about important cryptanalytic problems, the Albatross machine in particular, Friedman airily replied that NSA did not believe any outsiders could make a contribution to solving Albatross unless they were willing to devote months of continuous effort, which precluded the use of consultants such as themselves.45

  “There appears to be a kind of deadlock due to the fact that the Agency is reluctant to clear and brief a consultant for a high-level, compartmented program unless he agreed to come to the Agency and work on the problem for at least three months, and mathematicians are reluctant to commit themselves to at least three months on a problem about which they know almost nothing,” complained C
harles B. Tompkins, a SCAMP member and PhD mathematician who had also worked at ERA in its earliest years. In putting together SCAMP, Tompkins pointed out, every effort had been made to recruit the strongest possible group of cleared mathematicians, “and it is completely illogical to withhold from them a problem which is largely mathematical in character and which has defied solution for several years.” He added, with obvious frustration, “It is not true that mathematicians claim to be superior to cryptanalysts at cryptanalysis; they do claim to be better at mathematics”—precisely the point Friedman refused to acknowledge.46

  The best case that the NSA cryptanalysts could make for keeping the job within the agency was their faith—and it was becoming little more than faith—that the faster computers the agency was pouring tens of millions of dollars into acquiring would eventually achieve the long-sought breakthrough on the Russian problem. Largely because of NSA’s lavish patronage, the capabilities of commercially produced computers were increasing rapidly, with IBM, Sperry Rand, and other new companies introducing substantially more powerful models every year.*2 The first magnetic core memories, the first high-speed tape drives, the first all-transistor computer, the first desktop-sized computer, and the first remote workstation were all built in response to NSA orders, and the commercial markets they helped spawn stimulated further innovation. NSA took delivery in 1953 of an IBM 701, the world’s first commercially available scientific computer, which offered a tenfold jump in processing speed on standard cryptanalytic tasks over Atlas; the IBM 704 and 705 that followed in 1956 increased speed by another factor of five.47

  None of these general-purpose machines were specifically optimized for high-speed cryptanalytic mass data processing, however. Programming the early machines was a huge challenge. Programs had to be written in machine language, unique to each different computer; the simpler, higher-order programming languages like FORTRAN and BASIC did not yet exist. A slightly sardonic article in NSA’s in-house technical journal in October 1956 entitled “Computers—The Wailing Wall,” described the yawning culture gaps between the cryptanalyst attempting to devise an innovative and experimental attack on a recalcitrant cipher; the “methods analyst,” whose job it was to try to transform “vaguely expressed analytic ideas into exact logical procedures which can be represented on a machine”; and the programmer—who typically “holds a BS in mathematics, has been exposed to eight weeks of cryptanalysis in training school, and has been learning programming for six months”—whose job it was to make it happen.