Code Warriors Page 8
Cecil Phillips had made another astonishing discovery the following month. Some of the trade messages, he found, revealed another sort of reused one-time key: the Russian code clerks, when they had a long message to encipher, had adopted the insecure expedient of using a page of key in the normal fashion for the first fifty groups, then using the same key page in reverse order for the next fifty groups. A search for these reuses eventually yielded four thousand such “reverse depths.” The work was referred to as the “Red Reverse” problem, and a dozen staffers were assigned just to this task.8
It was now becoming clear that multiple pages of duplicate key had been issued by the Russians to users of the various one-time-pad systems; some of the duplicated pages were found in two different pads within the same system, but sometimes a page from a ZET pad also showed up in a ZDJ pad. As careless a blunder as it was to reuse any one-time-pad pages, however, none had been reused more than once: there might be two messages enciphered with the same key, never three or more. The conventional wisdom among cryptanalysts had been that having no more than pairs of overlapping messages to work with—a “depth of two”—was insufficient to break an enciphered code; it took multiple sets of overlapping messages each having depths of three or more to get anywhere. Arlington Hall’s Russian section had already proved that wrong, working entirely from depths of two.*2 By early 1946 some fifteen thousand groups of additive key had been recovered and about two thousand possible code groups had been identified in the ZDJ codebook (which Arlington Hall designated “Jade”), but half of those were considered “doubtful” and only twenty had been assigned even tentative meanings. The path forward was clear in theory, but agonizingly slow in practice: over the following months the section added only about ten code groups a month to the list of known words in the Jade codebook. If the standard work of a codebreaker was looking for the proverbial needle in a haystack, the Russian problem required finding a wisp of straw in a haystack.
In January 1946, Meredith Gardner, a linguist who had worked as a translator on the Japanese military attaché codes, joined the Russian section as its chief “book breaker,” whose job it was to work out the code group meanings. Gardner was a shy, slender, self-effacing scholar who had been teaching Spanish and German at the University of Akron; he had a master’s in German from the University of Texas, where he had also learned Russian by taking lessons from the Russian-born grandmother of a fellow student. He had similarly picked up Bulgarian, Hebrew, and other languages along the way through his own studies. After being sidetracked for part of 1946 to work on Bulgarian diplomatic traffic, Gardner was back at Arlington Hall’s Russian section on November 4.
Over the next two weeks he added eighty-six new recoveries to the Jade codebook, a burst of progress that nearly doubled, to two hundred, the number of known code group meanings. Then on December 13 he broke the Jade codebook’s spell table from a ZDJ message from 1944 that quoted a lengthy report in English. (The message itself was of no intelligence value, consisting of predictions on the likely outcome of the U.S. presidential election.) A week later Gardner broke part of another 1944 message, which also contained large stretches of spelled-out words: they were the names of “scientists who are working on the problem,” the message stated, and the list included Hans Bethe, Niels Bohr, Edward Teller, Enrico Fermi, Emilio Segrè, Arthur Compton, Ernest Lawrence, and Harold Urey—all leading scientists who were working on the Manhattan Project, a fact that in 1944 was Top Secret.
It was the first substantial result from the three-year effort to break the Soviet diplomatic codes. ZDJ, it was about to become all too obvious, was not handling the ordinary diplomatic messages of ambassadors and consular staff at all. It was, rather, nothing less than the primary communication channel for spies of the NKGB, the Soviet foreign intelligence service, later to become the KGB. These encrypted telegrams, sent under diplomatic cover, were in fact the principal means by which Soviet agents in the West received instructions from Moscow and filed their intelligence reports back.9
Standing over his shoulder as Gardner decoded the atomic scientist message was a coworker, William Weisband. A linguist adviser to the project, Weisband was always a bit mysterious about what he called his “exotic” background. He knew Russian fluently, and could read and speak Arabic. But he spoke English with only a barely discernible accent, had served in the U.S. Army as a translator and cryptologist in North Africa and Italy, and was a friendly, naturally gregarious man who moved easily around Arlington Hall, where he had been stationed since the end of the war.
Gardner did not think anything of it at the time that Weisband was the first person to learn of his break into the Soviet espionage traffic.10
—
The interception of Russian army, air force, navy, and police messages was meanwhile amassing a vast mountain of unprocessed material. By 1946, Arlington Hall was receiving from U.S. and British intercept stations twenty thousand messages a month of Russian military Morse code traffic. Dozens of different code systems were being studied and some were far enough along to be readable, but their intelligence value was slim at best. Nearly all used fairly simple hand encipherment systems such as additive key books that were heavily reused, and thus no great challenge in principle to break; or transposition ciphers in which the order of the figures in the encoded messages was scrambled according to a determined pattern, again a familiar exercise for the codebreakers by this point; or even extremely simple monoalphabetic substitution ciphers in which each letter was replaced by another in an unvarying pattern, which posed almost no real challenge at all. The question was whether the still sometimes tedious and manpower-intensive drudgery required in each case was worth it. Clearly none of these hand systems were carrying high-level communications. If the American codebreakers had been hoping to re-create the triumphs of the war, when they had penetrated the D-Day battle plans of Hitler’s generals or read the verbatim orders of Admiral Karl Dönitz to his U-boat wolf packs in the Atlantic, they were so far disappointed; instead of strategic assessments from the Red Army high command or secret orders from the Kremlin, they were reading a report on animal diseases in Siberia from an army veterinarian or an accounting of railcars under repair.
It was likely that the Russians were using machine-generated ciphers to protect their most secret military communications, but what those devices might be had been nearly a complete unknown to the American and British codebreakers during the war. West Coast intercept stations had begun to pick up radio teleprinter signals from the Soviet Far East in late 1944, however, and these seemed potentially important enough that Op-20-G ordered Lieutenant Tordella to drop what he was doing at Station S, attend a twelve-week course at Bell Laboratories in Manhattan on “special processing equipment,” and then head back west to set up and run the new experimental test station—designated Station T, it was located at Skaggs Island in San Pablo Bay, just north of San Francisco—that would attempt to apply the “special” equipment to recording the Russian signals.11
The interest sparked by the Soviet radio teleprinter traffic came directly from GC&CS’s phenomenal intelligence coup of breaking German military teleprinter signals. The German high command employed several teleprinter enciphering machines—the Lorenz SZ40/42 and the Siemens T52 Geheimschreiber—for their highest-level communications between Berlin and the headquarters of theater commanders and army groups. In one of the most mathematics-and machine-intensive cryptanalytic feats of the war, the Bletchley codebreakers had, without ever capturing or seeing one of the actual German machines, reconstructed the strings of additive cipher generated by these devices (which they called “Tunny” and “Sturgeon” respectively, “Fish” collectively) and developed an electronic special-purpose computer, the Colossus, to match intercepted messages with the correct string of additive key to break them. The messages read at Bletchley included many giving detailed German military plans, dispositions of forces, and orders before and after the Normandy landings. It was a not entirely forlor
n hope that the Russians might be using a similar teleprinter-based system to safeguard their most important military signals—and that it might prove equally vulnerable to a concerted attack.
Successfully intercepting the Russian radio teleprinter signals was a challenge to begin with, because the Russians, it was evident, were employing a complex transmission technique known as multiplexing, which combined two or more streams of traffic together in a single radio channel. Teleprinter machines used a system akin to Morse code, known as the Baudot code, which represented each keyboard character as a five-bit sequence made up of “marks” (conventionally denoted as X’s) or “spaces” (denoted as dots). The Baudot code thus allowed for 25, or 32, different possible characters. In the Russian version, Л, for example, was X X • X X, И was • • X • •, Д was X X X X •; the special character • • • X • triggered a carriage shift from letters to figures to allow some keys to do double duty as numerals, punctuation, and less frequently used letters of the Cyrillic alphabet. The teleprinter could also be operated by a paper tape punched in five rows according to the Baudot system, with a hole representing a mark and no hole for a space.
A five-bit paper tape punched with a message in Russian; for details of the Russian Baudot code and teleprinter encryption devices, see appendix B.
Once a message was prepared by punching it onto a tape, the entire transmission and reception process was fully automatic. There was none of the tapping out of each letter of a message by hand on a telegraph key while a radio operator at the other end transcribed onto a sheet of paper the dots and dashes he heard coming through his headphones; the paper tape simply was fed into a reader at one end, and the printer at the other end obediently clattered out the complete text. Sent over the airwaves, radio teleprinter signals consisted of a rapid, continuous stream of marks and spaces, with the marks usually represented by a small shift in the frequency of the transmitter’s signal. In multiplexing, two or more signals were interleaved in a single stream, like perfectly shuffling two decks of cards together. To separate the signals back out at the other end, a “demultiplexer” with a rotating distributor shunted each successive incoming bit onto a separate wire attached to its own printer.
Some of the Russian signals carried as many as nine multiplexed messages at a time, but those nine-channel signals appeared to be all unencrypted, plaintext traffic dealing mostly with commercial and economic matters. The encrypted military teleprinter traffic was carried by two-channel multiplexed signals, and that was what Tordella was told to focus on. Within a few months he had an experimental demultiplexer operating and was filing the separated-out streams of intercepted teleprinter signals back to Washington. (Only later did he learn that the Army was doing exactly the same thing at its nearby intercept station at Two Rock Ranch in Petaluma, California.)12
Keeping the demultiplexer’s distributor turning in exact synchronization with the incoming signal proved to be a maddeningly touchy business. The attitude of at least some of the higher-ups at Arlington Hall and Nebraska Avenue was that the whole thing was a phenomenal waste of time in any case, given the cryptanalytic challenges involved. One young Army officer who pressed for more attention to the effort remembered being told that there was no point in throwing away more time and money on a project that would only “add to the growing stack of unprocessed intercepts.”13
—
Help came from an unexpected quarter. A full year before the end of the war, GC&CS had been drawing up plans to send small teams of cryptologic experts along with Allied troops advancing through Germany to locate and seize the equipment and records and, if possible, the personnel of the German signals intelligence services. The aim was to discover any German successes that might be exploited, to learn if the Germans had broken any Allied codes and possibly shared that information with the Japanese, and to secure or destroy any documents that might expose sensitive cryptanalytic techniques. An unstated aim was to get to the German codebreaking experts before the Russians did. In August 1944, General Marshall approved the plan and asked General Eisenhower to accommodate the teams, which would include American cryptologists drawn mainly from the U.S. units already at Bletchley Park. The operation was given the opaque code name TICOM, which stood for the equally opaque Target Intelligence Committee.14
Throwing a bunch of very unmilitary civilian linguists and mathematicians in uniform into an active war zone to carry out a James Bond–like mission was undoubtedly a gamble. Art Levenson was a math major from City College who by the skin of his teeth had extricated himself from infantry training at Fort Dix after being called up, and managed to get himself assigned to Arlington Hall, and then to the Signal Corps contingent sent to Bletchley to work on the Enigma project with the British. He recalled the reaction of the regular-Army sergeant in their unit who “couldn’t get over this outfit” whose members could barely manage to appear with their uniforms on correctly: when the sergeant heard that Levenson and a few dozen others were going to Germany in April 1945, he said, “Boy, the war must really be over if they’re sending you guys.” Levenson was assigned to TICOM Team 1, led by a British major, John Tester, who had run one of the two groups at Bletchley that worked on Fish, and one of their key objectives was to try to capture an intact T52 Geheimschreiber. The senior American on Team 1 was Howard Campaigne, a Navy lieutenant commander who had been a mathematics professor at the University of Minnesota.15
Theirs was a frequently surreal adventure. A week after the German surrender the members of Team 1 were picking through Hitler’s, Goering’s, and Ribbentrop’s villas at the Nazi leaders’ alpine retreat at Berchtesgaden. The not very convincing cover story to explain their presence was that they were there to check that various German headquarters were obeying the terms of the capitulation forbidding the use of cipher equipment. Getting permission from local American commanders to enter restricted zones or carry equipment out was a constant challenge. Following a report that Field Marshal Albert Kesselring’s entire communication “train,” four trucks equipped with a T52 machine and radio gear, was located nearby at Zell am See, just south of Berchtesgaden, two members of the team took a jeep over a back mountain road, traversing rocks, meadows, and snowdrifts to get around a blown bridge and evading patrols of the 101st Airborne Division, which had issued a “freeze” order forbidding any movement in the area. When they arrived, the TICOM team found officers of the Luftwaffe high command headquarters still very much in charge: they were all armed, had posted sentries with submachine guns, and “seemed quite unconvinced that the war was irrevocably at an end.” Only after a considerable amount of persuading did the Germans reveal the whereabouts of the communications train, which the Americans were then able to secure.
Getting the captured trucks, and the enlisted men who had operated them, out of Germany produced other absurd moments. Levenson and Tester accompanied the convoy, the Germans driving their own vehicles. The men proved far more cooperative than their officers, genuinely eager to be of assistance and happy to have escaped falling into the hands of the Russians, so Levenson—a Jewish kid from Brooklyn—ended up letting one of his German prisoners carry his rifle for him. This led to a tense confrontation with an angry mob of civilians in Belgium, who, seeing obviously German trucks and an armed German soldier, thought the Nazis were back.16
The TICOM teams had few leads to go on and for the most part would simply show up at the headquarters of a U.S. Army unit and ask whether they knew of any German signals units or research stations in the area, or any prisoners they were holding who had worked in signals intelligence. A lot of leads went nowhere, and for the most part the occupation forces had just thrown a guard around any German installations they found without bothering to learn what they were. Campaigne remembered following up one report about a possible German research facility high in the Tyrolean mountains; driving up there, he found a lone GI standing guard at the door. Campaigne asked him what the Germans had used the site for. “Aw, them Krauts”—the soldi
er replied with a dismissive wave of his hand—“they built all kinds of shit.”17
On May 21, Campaigne received a much more solid tip. The U.S. Seventh Army was holding at the POW camp in Bad Aibling, between Berchtesgaden and Munich, a prisoner who had told the camp authorities that he had worked on intercepting and decoding Russian radio teleprinter traffic in a unit that had most recently been stationed at a barracks in nearby Rosenheim. Campaigne hastened over to interview the man. Unteroffizier Dietrich Suschowk told him that he and the other nineteen men of the unit—all now prisoners in the camp—had buried their equipment and documents on the grounds of the barracks. The next day Campaigne returned with a truck, collected the twenty prisoners, drove them to Rosenheim, and set them to work digging up their cache of cryptanalytic secrets. There was eight tons in all. The Germans were “helpful but afraid” at first, but quickly “showed the greatest willingness to cooperate”: the men, all NCOs, viewed themselves not really as soldiers but as “specialists, with a genuine pride in their work,” and when Campaigne evinced an interest in the technical aspects, they eagerly offered to reassemble one of the devices and demonstrate how it worked.18
On June 5 the equipment was on its way to England by air, and on June 29 six of the prisoners who seemed most knowledgeable about its operation arrived at the village of Steeple Clayton, fifteen miles southwest of Bletchley Park, where their gear awaited them. By the next evening they had one of their prize pieces of apparatus assembled and working. The Hartmehrfachfernschreiber, or HMFS, was a demultiplexing receiver that by inserting different distributors could separate out anywhere from two to nine multiplexed channels. It was an extremely sophisticated piece of equipment. An oscilloscope provided a visual indication of when the receiver was properly synchronized, and an automatic circuit then locked in the synchronization; a short-term memory buffer using relays automatically converted the rapid incoming pulses of the radio signals to the longer, 20-millisecond length that the teleprinter machines operated at. The following day the Germans had fifty teleprinters hooked up, printing out streams of intercepted Russian two-and nine-channel signals.