THE CODEBREAKERS

by DAVID KAHN

  Even though the alphabet card on which the Larrabee was sent to American diplomatic and consular offices, with instructions to paste it inside the cover of departmental codebooks, would not help much in solution, one nation was taking no chances. Vice Consul General Alfred V. Smith wrote on September 13, 1913: “In reference to Circular Instruction ‘Larrabee Cipher’ of March 8, 1913, received at this office on April 3rd last, I have the honor to inform the Department that the ‘Larrabee Cipher Code’ referred to in the above instruction as being ‘transmitted herewith,’ was not to be found in the envelope.” Smith was in charge of the consulate general at Moscow.

  In October it was learned that there were no funds available for the proposed interdepartmental code, and so the United States government persisted in the amazing cryptographic imbecility of using the Larrabee throughout World War I. For strictly diplomatic messages, State continued to use RED and BLUE, relying increasingly on the latter as the former declined in security. In 1915, Mexicans somehow obtained the RED code at Vera Cruz. Across the Atlantic, the American minister to Rumania, a former politician, found it easier to keep his one copy under his mattress than to fiddle with the safe-combination. It disappeared one day, reportedly finding its way to St. Petersburg. The minister never troubled to report this loss. He solved the potentially embarrassing problem of not being able to read incoming messages by letting the relatively light coded traffic pile up, then hopping a train to visit the embassy in Vienna, where he decoded the messages which, he said, had arrived just as he was leaving. At the same time he composed and encoded his replies. When the war came, however, traffic increased, the artifice no longer applied, and he admitted his dereliction and returned to politics.

  The need for security penetrated the departmental bureaucracy and in 1914 it instituted a “special cipher,” probably a superencipherment of the existing codes. The London embassy gleefully hailed it as a success and an ensurer of secrecy. It was probably only a stopgap until the State Department could compile and place in service a new code—the GREEN. This one-part code used five-letter codegroups of the form CUCUC (C = consonant and U = vowel); thus department was FYTIG, message was MIHAK, secured from was PEDEK, secured the, PEDIV.

  Nevertheless, by the time the United States entered the war, every major European power must have had copies of one or more American diplomatic codes. Foreign employees had the run of the embassies, and it would have been little trouble for them to get hold of the books. (One of the first German spies arrested was the clerk to the commandant at Pearl Harbor, who had access to the Navy’s most secret code.) The Germans even returned a State Department codebook that had been used by the American consulate in Leipzig. If England’s Room 40 could solve German two-part codes, it could certainly read the simpler American ones, and in fact rumors that Britain was doing so even reached the newspapers. During the war, the London embassy reported that German authorities in Spain obtained a copy of an American cablegram to Valencia, radioed the codetext to Germany, and promptly got back the plaintext. The State Department attempted to counter such embarrassments by enciphering the GREEN code with a key that changed monthly and by producing a new code (“I never realized until now what an arduous task it is to prepare a wholly new cipher,” sighed Under Secretary William Phillips.)

  In 1919, the United States, which Will Rogers said never lost a war nor won a peace, may have been assisted in losing the current peace by its cryptographic practices. The American Commission to Negotiate Peace furnished its field agents, who reported to it on the conditions and aspirations of the little peoples of Europe, with a publicly available commercial code, the Universal Pocket Code! No doubt the French cryptanalysts were pleased to see their work thus facilitated.

  The new code that Phillips had sighed over was the GRAY code, destined to become the best known and longest lived of American diplomatic codes. In theory it was a confidential code: when a telegram from Mexico reported a rumor that that country had obtained an American code (not the same as the RED stolen earlier), the department replied that GRAY could be used for confidential messages. But a truer picture might be that drawn by former code clerk James Thurber, who served with the peace mission:

  All our code books except one were quaint transparencies dating back to the time when Hamilton Fish was Secretary of State under President Grant, and they were intended to save words and cut telegraph costs, not to fool anybody. The new code book had been put together so hastily that the word “America” was left out, and code groups so closely paralleled true readings that LOVVE, for example, was the symbol for “love.”

  Whatever slight illusion of secrecy we code clerks may have had was dispelled one day by a dour gentleman who announced that the Germans had all our codes. It was said that the Germans now and then got messages through to Washington taunting us about our childish ciphers, and suggesting on one occasion that our clumsy device of combining two codes, in a desperate effort at deception, would have been a little harder if we had used two other codes, which they named. This may have been rumor or legend, like the story, current at the time, that six of our code books were missing and that a seventh, neatly wrapped, firmly tied, and accompanied by a courteous note, had been returned to one or another of our embassies by the Japanese, either because they had finished with it or because they already had one.

  A system of deception as easy to see through as the passing attack of a grammar-school football team naturally produces a cat’s-out-of-the-bag attitude. In enciphering messages in one code, in which the symbol for “quote” was (to make up a group) ZOXIL, we were permitted to use UNZOXIL for “unquote,” an aid to perspicuity that gave us code clerks the depressing feeling that our tedious work was merely an exercise in block lettering. The Department may have comforted itself with the knowledge that even the most ingenious and complex codes could have been broken down by enemy cipher experts. Unzoxilation just made it a little easier for them.

  So did the continued use of the Larrabee. In 1921 the Navy awoke to the danger. Commander Milo F. Draemel conferred with Salmon, and the two agreed that a double transposition cipher would afford sufficient security for the limited State-Navy communications. But not until almost a year later did the State Department finally distribute it to 16 legations and 59 consulates around the globe.

  What also made the work of foreign cryptanalysts easy was America’s continued use, year after year, of the old codes. The GRAY code especially became so familiar to American foreign service officers that when colleagues tendered a senior consul at Shanghai his retirement dinner late in the decade, he responded with a farewell speech in GRAY—which the old-timers followed with ease. By then, superenciphered codes called A-1 and B-1 had been introduced. In 1925, the chargé d’affaires at San Salvador, pointing out that the GRAY and the GREEN codes were too old to be secret but that they were still being used for confidential messages, suggested that the department mandate the A-1 for ordinary confidential messages and the newer B-1 for highly confidential ones. The department seems not to have adopted this sensible suggestion. Instead it continued to use what Herbert Yardley, then its quasi employee, bitterly and accurately called sixteenth-century codes.

  From time to time Salmon’s office would test a new system, as it did with the Hitt cipher machine that Friedman so promptly broke. But more often Salmon would send out—without even examining the system—a stock, smug reply to each of the dozens of proposals that poured in on him from inventors: “The Department is in receipt of your letter of the 25th instant making reference to the creation by you of a new code system and inquiring if there is a usage for such a code. I beg to inform you in reply that the codes and ciphers now in use are adequate to the present needs of this Department.”

  Why did the Department of State not introduce improved methods of cryptography—perhaps the Vernam machine or the one-time pad—at least for its more secret messages? Apparently nothing more than bureaucratic inertia, probably compounded with some budgetary tightness. As Yar
dley said: “There is only one indecipherable means of communication, and its adoption would require the Department to revolutionize its antiquated methods.” His conference in the late 1920s with a high department official, who had summoned him to discuss the problem because the official had heard that Mexico was reading American code messages, ended on that note.

  Nothing less than an international scandal would wake up the government to the fact that the very basis of all successful diplomacy is safe and secret lines of communication. But my whole life had been devoted to destruction. I should like to leave a monument to constructive cryptography.

  As I walked through the wide high corridors on my way to the entrance, I mused how proud one might be to leave to the United States Government a method of communication that would insure the secrecy of her dispatches throughout the ages. Aside from this, of course, was professional pride. Then too it would be fun to laugh at foreign cryptographers as in my mind I saw them puzzling over our secret telegrams, striving in vain for a solution.

  But why dream? After all, weren’t all diplomatic representatives just funny little characters on a stage, whispering, whispering, then yelling their secrets to the heavens as they put them on the cables!

  American cablegrams certainly yelled. In 1929, Charles G. Dawes, ambassador to the Court of St. James’s, found it necessary to wire the State Department: “Suggest telegrams contents of which under instruction are to be conveyed to the British Government should not be coded in such confidential code.” In 1931, Stanley K. Hornbeck of the Division of Far Eastern Affairs minuted to Secretary of State Stimson: “Mr. Secretary: I have the feeling that it is altogether probable that the Japanese are ‘breaking’ every confidential telegram that goes to and from us, in Japan and in territory controlled by Japan. It is not impossible, but less likely, that the Chinese are doing the same thing. Whatever may be the facts, I feel that we should have the possibility constantly in mind.” A few months later, as if in confirmation, the chargé d’affaires at Tokyo reported that the American consul general at Seoul, Korea, then a part of the Japanese Empire, had received new A-1 cipher tables with wax seals broken. By this time, the insecurity of American codes seemed almost to be taken for granted. “I could not help wondering,” Ambassador to Japan W. Cameron Forbes cabled Stimson on February 16, 1932, “whether in view of the imperfection of our codes, this [aforementioned dispatch] might not be read off by the Chinese secret service people”—a reading which, he thought, “would immediately put the Chinese against the proposition.”

  This disquieting state of affairs remained pretty much in effect during the thirties. State basically continued to use the same old codes. Ambassadors still had to decode especially secret messages intended for them alone. For at least one, Harry F. Guggenheim in Cuba, the excitement of knowing that the message would be important mitigated the onerousness of the task. Under the prodding of President Roosevelt, who had learned to pay attention to communications security as an Assistant Secretary of the Navy in World War I, State produced several new codes, among them the BROWN. The BROWN was stolen from the American consulate at Zagreb by a gang of Ustachi bandits as a by-product of a safe-cracking raid, but it continued to be used elsewhere in the world. In addition, C-1 and D-1 codes had been added, each—like the A-1 and B-1—with its own tables for superencipherment. These tables served for irregular periods ranging from two to five months. They added slightly to the security of the codes but greatly to their cumbersome-ness: during the Munich crisis, half the vice consuls in Berlin had to work in the code room.

  Late in the 1930s, the department adapted as its most secret method of cryptography the system invented a century and a half earlier by the first Secretary of State, Thomas Jefferson. This was the strip cipher, the M-138, the flattened-out Jefferson wheel cypher-Bazeries cylinder, first constructed in strip form by Parker Hitt in 1914. Each M-138 set had 100 strips. In enciphering, 30 at a time were used, with the ciphertext being read off, not from one generatrix, but from two in groups of 15 letters each. So long as the strips were kept secret and were changed often enough, the strip system apparently secured important American diplomatic correspondence. State used it, for example, to encipher a triple priority message from Roosevelt to Churchill just after the Atlantic Conference.

  President Roosevelt, however, distrusted State Department codes on principle. So when war broke out in Europe, he communicated with his ambassadors in London, Paris, and Moscow via Navy Department crypto-systems for “matters of utmost secrecy.” This annoyed the State Department, which felt that diplomatic matters were being kept from it. “But,” William C. Bullitt, ambassador to France, later wrote, “I should regret to have the impression prevail that this was due to a desire to conceal anything from the Secretary of State, Mr. Hull, for whom I always have had high respect. It was due to lack of security in the Department of State. The codes of the department were so antiquated that, in the opinion of the President, they had been broken by all the totalitarian powers.” Admiral W. H. Standley, ambassador to Moscow, said that “It was also common gossip that the State Department codes were insecure.” The American embassy in Madrid received its most secret instructions in a British code!

  Both Bullitt and Standley also mentioned leaks in the State Department as another reason for using Navy codes. Leaks there were—and from the very heart of State’s communications. On April 30, 1940, Hans Thomsen, Germany’s chargé d’affaires at Washington, cabled home: “A reliable and tried confidential agent who is very friendly with the director of the code room of the State Department reports as follows after having seen the relevant telegraphic reports.” On September 30, Thomsen relayed to Germany intelligence that the same reliable informant had obtained from a cablegram sent to Roosevelt by the American ambassador to Britain, Joseph P. Kennedy. On December 29, the German ambassador to Spain wired home about “a code telegram from Cordell Hull to the United States ambassador here on December 18, the text of which has become known to me.” All these helped Germany in shaping her foreign policy. The message to Madrid, for example, showed that the Spanish Foreign Minister was lying to Germany when he insisted that Spain would not receive needed grain if she stayed neutral.

  The worst of the leaks took place in London, where Tyler Kent, a bright, handsome, but twisted young man, worked in the code room of the American embassy. Convinced that a vast Jewish conspiracy was pushing the United States into an unwanted war, and that to help the enemy of the Jews was to help his own country, Kent took telegrams from the embassy and passed them to a pro-Nazi group. By various channels they reached Germany. The German ambassador in Italy wired home a report of Roosevelt’s reply to Churchill’s request for 50 destroyers only seven days after it was received in London. On May 20, however, Scotland Yard plugged the leak. It arrested members of the pro-Nazi group for espionage and, with State Department approval, searched Kent’s rooms.

  There police agents found copies of more than 1,500 embassy papers, many of them telegrams, as well as two newly made duplicate keys to the index bureau and the code room of the embassy. When the stunned ambassador asked him why he would betray his country this way, Kent explained that giving the documents to Germany would help keep America out of war. He was instantly dismissed, and then, having lost his diplomatic immunity, was arrested, tried, and convicted by the British for violating the Official Secrets Act and sentenced to seven years in prison.

  But the damage had been done. “The removal of so large a number of documents from the Embassy premises,” the State Department later declared, “compromised the whole confidential communications system of the United States, bringing into question the security of the secret ciphers.” Kennedy said: “Because of his [Kent’s] treachery, all diplomatic communications of the American diplomatic service were blacked-out at a most critical moment in history—during the days of Dunkirk and the fall of France. The blackout, which concerned the American embassies and missions throughout the world, lasted from two to six weeks until scor
es of special couriers had reached the embassies with new codes from Washington.”

  On the other side of the globe, meanwhile, poor American cryptography gave aid to a potential enemy and injured the cause of peace. Within the cable section of Japan’s Foreign Ministry was hidden a small cryptanalytic group, the Ango Kenkyu Han (“Code Research Section”). About five of its members worked on English and American codes. Each morning a messenger from the Foreign Ministry (and others from the Army and Navy) picked up copies of foreign diplomatic telegrams from the Communications Ministry’s censorship department. Of the main codes then in use by the American embassy—GRAY, BROWN, A-1, B-1, C-1, D-1, and M-138—the Ango Kenkyu Han could read three or four of the lower grade. Ambassador Joseph C. Grew may have inadvertently helped them. “One of the high officials of the Japanese Government wanted to send a secret message to our Government which they did not want the Japanese military to see and in passing this message on they asked me to please put it in our most secret code. I said of course I would do so.” Even so, the Japanese failed to penetrate the M-138 and the higher-grade codes.