Code Warriors
Page 15
It also was designed to carry out a number of special functions unique to cryptanalysis, including adding alphanumeric characters together in modular arithmetic (as was involved in adding or subtracting key-in systems such as the Russian teleprinter ciphers and the Hagelin) and, even more important for cryptanalysis, initiating a running comparison of two data streams and counting coincidences between the two. The group had correctly noted that this fundamental cryptanalytic task was an inefficient one when carried out by the standard general-purpose computer program. Comparison runs involved performing exactly the same steps over and over; but the stored-program architecture’s very flexibility was a curse here, since it required the computer to retrieve and execute, one at a time, the same sequence of instructions for every single comparison of two individual numbers. Abner’s “Swish” function did away with all that by throwing the computer into a streaming mode that just pointed to two stacks of data and sent them running sequentially into an analytic unit that tallied coincidences without any further instructions; it then automatically placed one of the streams back into the memory a specified number of words away from its original location, ready to be retrieved again, thereby sliding the two sets of data through every possible offset.21
In fact, the war-era optical comparators would remain faster than the conventional general-purpose computer at performing this central cryptanalytic task until the third generation of digital electronics arrived in the mid-1950s.22 The electronic realization of the streaming function that Snyder’s group incorporated into Abner would be a key feature in the future mammoth, technology-pushing computers built for NSA by IBM into the 1960s.23 Only with the advent of supercomputers would the advantages of this kind of specialized cryptanalytic computer architecture be overtaken by brute processing speed.
Atlas did not have a streaming function, but ERA, at Op-20-G’s request, built a number of specialized electronic comparators that did much the same thing. The first of them, Goldberg (ERA’s “Task 9”), cost $250,000, could perform twenty thousand comparisons a second, and did all the basic cryptanalytic tests, including round robins, index of coincidence counts, crib dragging, and language-weighted statistics. A more specialized machine, Demon, was rushed into production in the meanwhile to tackle one of the Soviet high-echelon teleprinter encryption systems; equipped with a magnetic drum memory taken from Goldberg, it was designed to test a large number of cribs at every location of one of two messages already found to be in depth, and determine whether the resulting plaintext in the paired message yielded recognizable high-frequency Russian words stored in a dictionary.24 The first Demon was delivered in October 1948, just in time for the whole Soviet problem to come crashing down.
—
On June 24, 1948, the Soviet military administration in Germany, without warning, issued orders halting all rail traffic into the western half of Berlin. The blockade left the two and a half million German civilians living in the French, American, and British sectors of the former capital completely cut off from the supplies of food and coal that had kept them alive since the end of the war. Lucius Clay, the military governor of the American zone, later called it “one of the most ruthless efforts in modern times to use mass starvation for political coercion.” It was Stalin’s first move from diplomatic obstructionism to outright military confrontation with the United States.25
Exactly what Stalin hoped to gain remains uncertain to this day. The immediate precipitating event was a plan announced by Clay to introduce the new West German deutschmark in West Berlin to replace the inflated and worthless notes the Russians had flooded Germany with, and Stalin may have been trying to keep the West marks out. He may also have been more directly hoping to force the Western allies to abandon their military presence in Berlin altogether, or pressure them into backing away from their moves to end the military occupation in the western zones and establish a new German state. Berlin was in any case the place where the Soviets were confident they could apply the maximal pressure with the least cost to themselves, and so it would remain throughout the Cold War. (As Khrushchev, with his usual earthy directness, would later remark, “Berlin is the West’s balls. Whenever I want to make the West scream, I squeeze Berlin.”)26
Four days later Truman approved Clay’s ad hoc decision to respond to the blockade by trying to supply Berlin by air. Recognizing the enormous political risks, but believing that basic principles were at stake—as well as any hopes for the formation of a West German government that could anchor the stability of postwar Western Europe—Truman did not even consult with advisers before issuing his instructions: “We stay in Berlin, period.” A month later he overrode objections from his Air Force generals and ordered 160 large four-engine C-54 cargo planes, more than half the force’s entire global airlift capacity, to join the operation, then increased that to 200 in early September. No one had thought it possible to supply by air forty-five hundred tons of food and coal a day, the amount Clay calculated was the bare minimum needed to keep the city from starving and its electricity plants running, but C-54s were landing every couple of minutes at Tempelhof Airport and the city’s other two tiny airports, idling on the tarmac with two engines running as crews hustled off ten tons of bulging burlap sacks, then were back in the air in less than twenty minutes.27
Truman took one other highly visible action, ordering two squadrons of B-29s, sixty of the mighty long-range bombers that, everyone in the world knew, had dropped the atomic bombs on Hiroshima and Nagasaki, to RAF airfields in East Anglia. Not made public was that the aircraft sent to Europe were not in fact equipped to deliver atomic weapons. Truman had impatiently cut short Secretary of the Air Force Stuart Symington at a meeting at the White House when Symington glibly tried to argue that custody of atomic bombs should be transferred from civilian control to the Joint Chiefs. “Our fellas need to get used to handling it,” Symington foolishly insisted. Truman shot back, “This is no time to be juggling an atom bomb around.” But the deployment of the B-29s was intended as a not very veiled warning to the Soviets, the first gambit in a grim game of nuclear bluff that both sides would play with increasing realism and risk as the Cold War progressed.28
The cryptanalysts at Arlington Hall and Nebraska Avenue were absorbed in their own crisis while the Berlin confrontation played out in public. For over a year they had been watching with growing unease a series of changes in Soviet communications security practices. In August 1946 the Russians began enciphering the indicator groups that told the intended recipient of an enciphered code which key pad or additive book page had been used. The change occurred in all of the five-digit codes used by the armed forces, the MVD, and the MGB. In September 1947 there was a noticeable drop in the number of official messages transmitted in the clear on the internal radiotelegraph networks of the Soviet Union, and in December 1947 a notice was issued by the telegraph authorities in Moscow that coded telegraph messages were no longer to be sent over unscrambled radio teleprinter channels, but were restricted to landlines only; at the same time there was an upsurge in scrambled radio teleprinter traffic on certain links. In early 1948 military traffic in the Soviet Far East sent using the Sauterne machine, the Soviets’ version of the Hagelin B-211, went off the air; in April 1948, use of the Longfellow teleprinter encryption machine ceased. Throughout 1948, codebooks for a number of Russian enciphered codes, including some naval and MVD systems, were changed from one-part codes to the much more difficult to break two-part codes.29
None of these changes in themselves stood out from the sort of routine upgrades that happened all the time, but the tempo of the activity suggested an overall tightening of security. On Monday, November 1, 1948, something the U.S. and British codebreakers had never seen before took place. An urgent report the next day from the Op-20-G traffic analysis section cataloged the sweeping changes that had occurred in virtually every communications system of the Soviet military and MVD:
Unprecedented Coordinated Russian Communication Changes.
1.
Beginning 1 Nov extensive communication changes, which overshadow all previously recorded changes in type and areas affected, were effected in the Russian Naval, Military, and Police Communications Networks. Major fixed radio stations of all services (Navy, including Naval Air; Military, including Military Air; and Police) are now employing what appears to be the same type of call sign system, similar in appearance to International Berne call signs. Other sweeping changes occurred down the line of the three services affecting radio procedures, call signs, message formats, and in some instances the frequency plans.
2. Points of significance which arise from these changes include:
a) The efficiency with which the changes were executed in all areas by the three services, which is in direct contrast to previously recorded changes of individual services.
b) The coordination exemplified by the three services in enacting the changes.
c) The extensiveness of area application: All Naval Fleet areas; European and Far Eastern Police; and European and Far Eastern Military areas.
3. It is not evident that there was any reason for the execution of these changes on 1 Nov other than readiness by all services to do so. The coordinated timing does indicate clearly a central unified control or direction of all service communications.30
The Soviets had pulled the plug even more emphatically than was first apparent. All traffic passing over military, naval, and police radio links was replaced with nothing but practice and dummy messages. The Soviets’ heavy use of radio teleprinter and other radio links for their internal communications had been a recourse of necessity given the extensive damage to landlines during the war (and the Soviet Far East and Central Asia had never been well served by landlines and had always depended heavily on internal high-frequency radio networks). Reconstruction remained far from complete in late 1948. But apparently the Russians had concluded that their security concerns could not wait, and were shifting whatever they could of their most secret communications to available landlines immediately, even if it meant drastically cutting back the amount of traffic that could be handled.31
The November 1 change would become a fabled part of NSA’s lore, known within the agency as “Black Friday” (although the changes had actually been made over the weekend of October 30–31 and introduced the following Monday).32 The full extent of the disaster only became apparent the following spring when real traffic started reappearing on the radio nets, now employing greatly improved—and completely unbreakable—technical and security procedures. The keying errors or other mistakes that had allowed most of the Soviets’ machine-enciphered military traffic to be routinely read by U.S. and British codebreakers for the past several years had been corrected, and the much more disciplined systems that now replaced them slammed the cryptanalytic door shut.
The one important Soviet machine system that had remained on the air immediately following Black Friday was Albatross. When Coleridge traffic vanished on Black Friday, following the earlier disappearance of the Sauterne and Longfellow machines, some of that traffic was temporarily taken over by Albatross. But Albatross had stymied the best efforts of Arlington Hall’s codebreakers, and would continue to do so for years to come even as the cryptanalysts marshaled ever more powerful special-purpose analyzers, and then the first digital electronic computers, to attack the problem.33
The Soviet one-time-pad systems used for communications that passed through diplomatic cables also underwent a comprehensive alteration in indicators and addresses. Although Arlington Hall had not found any duplicate key use on messages sent after 1945 on most of these nets, it had continued to collect, sort, and study current traffic in the hope that reuses might still appear. (Some NKGB/MGB messages sent between Canberra and Moscow had employed the old duplicate key pad pages as late as 1948.) But with the indicator changes, it became impossible even to sort the traffic into the five different systems that had been identified: GRU, naval GRU, trade, consular, and MGB. Cecil Phillips was finally able to break the new indicators to allow that sorting and traffic analysis to resume, even if the messages themselves remained unreadable.34
A sudden, across-the-board change in Soviet code systems and communications procedures was so alarming that London and Washington briefly considered the possibility that it indicated preparations for imminent war. At a meeting in November 1948, the U.S. Communications Intelligence Board reported that their British counterparts had definitely ruled out that explanation but could offer only speculation in its place. The changes might be part of a “methodical drive to improve communication security”; they might be a temporary stopgap while security defects that had been discovered were corrected; or they might have been an urgent response to leaks about U.S. and British successes in reading Soviet cryptographic systems. The chiefs of the Army Security Agency and the Navy’s CSAW were asked for their views: the Army’s Colonel Hayes was “strongly inclined toward the belief that leakage of information had been the primary cause,” while Captain Wenger of the Navy leaned toward the more mundane explanation of “methodical” improvement, but acknowledged that a leak could not be ruled out. By the end of the year the board agreed to proceed on the assumption that the Soviets had been tipped off that their codes had been penetrated. USCIB ordered the “need to know” rule tightened and directed that even those specially cleared to receive communications intelligence at “user” agencies such as CIA were not entitled to know anything of the technical details concerning how it was produced. In April 1949, GCHQ and USCIB agreed that henceforth there would be a “complete separation of work on non-Russian from that on Russian,” both on the part of the traffic analysts, cryptanalysts, and translators who produced the material and the users who evaluated and disseminated the results for intelligence.35
Those were all prudent steps in any case, but they did nothing to reverse the damage that had already been done; nor were the British and Americans any closer to finding out if they did in fact have a Soviet spy or spies inside their signals intelligence agencies, and if so just who he or they might be.
—
At the start of 1949, 1,073 of the 3,124 workers at ASA and CSAW were assigned to the Russian problem; at GCHQ there were another 389. Hundreds more manned the 524 intercept radios tasked to collect Soviet radio traffic at the 38 field sites the United States operated around the globe. It had been a huge surge of personnel from the handful that had begun work on the Russian problem a few years before. Now it was not clear if any of them had a job to do anymore.36 On March 24, 1949, Frank Rowlett, now chief of ASA’s Operation Division, responsible for all communications intelligence (COMINT) production, and Solly Kullback, chief of its Research and Development Division, reported the one glimmer of hope left in the effort to exploit Soviet communications:
Russian communications security measures introduced over the period of the last eighteen months, including retirement of certain major cryptographic systems and the virtual cessation of operational radio activity on Armed Forces links, have increased the dependence of the allied Comint effort on the study and analysis of the large volume of traffic passed on the Russian internal civil radio links. These links, as yet unaffected by Russian communications security measures, are now the major source of current economic and current military intelligence information.37
These hundreds of thousands of routine, unenciphered telegrams carried each month over the Soviets’ civil internal radio networks had not seemed like a very promising source at first, and more than a few of the traditional-minded cryptanalysts in Britain and the United States, Rowlett among them, were initially dismissive of the idea that the Russians were going to hand over anything of intelligence value on a plate that way. “If the Soviet Union considered it important,” Jacob Gurin, a Russian linguist at Arlington Hall, remembered being told, “they would encipher it.”
But even before Black Friday a small plain-language unit under Gurin’s enthusiastic direction, drawing entirely on Russian telegrams that had been sent by radio in the clear, produce
d several insightful reports on the Soviet ministries in charge of industrial production. Gurin, known as Jack, spoke Russian fluently; born in Odessa to Russian Jewish parents, he had come with his family to America as a young boy but grew up speaking the language at home. Gurin argued that even the most tedious telegrams dealing with coal supplies, railcar loadings, and labor requirements could, when pieced together, produce a comprehensive picture of the state of the Soviet economy. In fact, given the near-total secrecy Stalin imposed on publication of official information (the locations of government ministries and production sites, the full names of officials, even texts of government decrees were often a secret) and the all-pervasive authority of the Soviet government in the country’s centrally planned economy, the plain-language messages represented a source of information available nowhere else.38
Gurin’s group had begun work in November 1947 with a staff of six but added nearly one hundred Russian linguists during 1948, many shifted to the effort when the encrypted B-211 traffic went off the air. By the summer of 1948, USCIB and GCHQ concluded a formal arrangement—it was embodied in a voluminous appendix K to the 1946 BRUSA agreement—extending their cooperation into Russian plain-language teleprinter and radiotelephone material, with the British contributing to the joint effort a special linguistic unit it had been operating since the end of the war: located on the third and fourth floors of a converted apartment building on Ryder Street, in the fashionable St. James’s district of central London, it was staffed mostly by Russian émigrés. Working from clues as tenuous as a list of Gosbank account numbers that the analysts were able to link to Soviet defense industries, Gurin’s group issued reports identifying centers of munitions production, assessing the capacity of the Soviet transportation system, estimating the output of vehicle assembly and engine plants, and compiling basic production statistics for steel, chemicals, oil, and electric power. For several years, the plain-language effort would also be one of the primary sources of information about the Soviet atomic program, and one of the few means to monitor warning signs that might indicate mobilization for war.39