THE CODEBREAKERS

by DAVID KAHN

  The unit worked near the front lines so as to get as many intercepts as possible. So close were they that on Leyte late in 1944 Japanese paratroops dropped on the unit, apparently having mistaken it for a command post because of its numerous antennae. One startled radioman, isolated in a direction-finding booth in the middle of a clearing, suddenly heard bullets whizzing all around him. The codebreakers dropped their pencils, grabbed their rifles, and engaged in rather more direct action against the enemy than that to which they were accustomed. The paratroopers were driven off, but not quickly enough to save the unit’s documents from the flames.

  Its radio operators, specially trained in Japanese Morse, listened in 24 hours a day on at least some of its two dozen receivers. Sometimes just the circuits being used would give Japanese intentions away. On Biak in 1944, the unit quickly learned that messages on a certain frequency invariably preceded an evening air raid—a bit of foreknowledge that enabled one member to collect regularly on sure-fire bets with a sergeant from a nearby outfit. Other times the 20-odd nisei in the unit intercepted Japanese cleartext. Usually, however, the radiomen typed out the coded intercepts and handed them to a traffic analyst. Most of the messages reported planes flying from one point to another, and the analyst, by a study of call-signs, could tell which unit and which points were meant. The 15 cryptanalysts had the mechanical task of stripping the additive from codes that had been solved at C.B. Each day a report summing up the unit’s conclusions went rearward to 5th Air Force headquarters, to which the unit had been attached, switching from the Signal Corps under C.B. to the Army Air Corps and receiving the new name of 1st Radio Squadron, Mobile.

  Success usually came in the humble form of an early warning of an air raid that probably saved American lives, or as some insight into a Japanese move that enabled an American commander to neutralize it. Late in the war, the unit’s solution of Japanese meteorological codes told American bombing commands what they wanted to know most—weather conditions over target. The outfit alerted the Allies to a major Japanese build-up when it solved a message reporting the presence in an airplane of two high-ranking officers of Japan’s 4th Air Army, which up to that time had been thought to be in northern China. But its greatest feat was the discovery of a huge concentration of Japanese air strength at Hollandia. The 5th Air Force launched massive raids and destroyed more than 100 enemy planes. Consequently they were not present to attack the American invaders, who splashed ashore with virtually no opposition.

  The Imperial Japanese Navy had commenced its cryptanalytic efforts in 1925 with the creation of an ultra-secret Tokumu Han (“Special Section”) in the 4th, or communications, Department of the Naval General Staff. It then numbered six persons, including clerks, and was located in the red brick Navy Ministry building in Tokyo. Among its early members were the young naval officer Hideya Morikawa, nephew of Chief of Staff Admiral Kanji Kato, and Morikawa’s former superior, First Lieutenant Kamisugi, who had handled cryptography aboard the flagship Nagato. Captain Kowalefsky, the Polish cryptologist who had improved the codes that Yardley had solved, lectured on cryptanalysis, and the neophyte codebreakers cut their eyeteeth on the GRAY code of the U.S. Department of State, making their entry through the classic technique of identifying NADED as period.*

  They also solved Chinese cryptograms during the Manchuria incident, primarily because these were based on a commercial codebook that transformed the Chinese ideographs to four-digit numbers for telegraphic communication. After the Japanese seizure of Shanghai early in 1932, Morikawa was sent there as chief of a cryptanalytic unit attached to the 3rd Fleet. He solved a Chinese message that corroborated a slightly doubtful Tokumu Han solution of an American GRAY message reporting Chinese plans to use its Air Force to attack Japanese troops. Instead the Japanese struck first, catching most of Chiang Kai-shek’s Air Force at Hangchow.

  The Tokumu Han failed, however, to break two-part codes, such as the State Department’s BROWN code, those used by the American Navy, and those introduced by Yardley into Chinese communications when he was Chiang’s cryptologist—except in extraordinarily favorable circumstances. One such occurred on February 26, 1936, when two regiments mutinied in Tokyo and several statesmen were assassinated in an attempted coup d’état. This furnished the cryptanalysts with an ocean of text and plenty of probable words to go fishing with. For a short time they read most American communications, including those of the naval attaché. Then the United States changed systems, and the skill of the Tokumu Han again proved unequal to its task. Its resourcefulness made up for this: near the end of 1937, Morikawa, accompanied by a locksmith, a photographer, and some lookouts, broke into the American consulate at Kobe and photographed the BROWN code and the M-138 cipher device, which the Japanese had never seen before.

  Soon thereafter, as part of Japan’s preparation for war, the naval shoguns built their first big intercepting post at Owada, a village about fifty minutes by car from Tokyo. During American naval maneuvers, its direction-finding and traffic analysis helped the general staff analyze American forces and tactics. The Tokumu Han also added cryptanalysts, all of whom were officers. By Pearl Harbor there were ten working full time and ten part time. They had still not succeeded, however, in reading American cryptograms.

  After Pearl Harbor, the rampant growth of Allied communications compelled the Tokumu Han to expand still further. The first batch of recruits—60 of them—were drawn from foreign-language schools and commercial colleges to become the first civilians in the Tokumu Han. The second batch consisted of about 70 reserve officer candidates selected from about 500 in basic training on the basis of their competence in foreign languages. (These signal intelligence groups differed from classes learning cryptography.) During a five-month course at the Naval Communication School at Kurihama near Yokosuka—hard by the Commodore Matthew Perry monument—they practiced International Morse, studied the elementary Oriental Tenji and Tenchi ciphers as well as the Occident’s more advanced Porta and Vigenère, and learned how to break codes and ciphers. Six classes, each larger than its predecessor, were trained during the war. Some graduates were assigned to communications intelligence in the intelligence units of fleet and force headquarters. In November of 1943, for example, the 3rd Fleet employed three officers and six enlisted men to monitor enemy messages. But most went straight into the Tokumu Han proper.

  A torrent of intercepts was pouring into it. Most came from the hundreds of radio receivers and direction-finders of the Owada Communications Unit. Some were picked up by the 20 Americans and Australians pressed into service with the Kanagawa Communication Force near Hiyoshi, and a few messages trickled in from fleet radio units. Near the end of the war a unit was set up in a radish field at Yokosuka. The entire Tokumu Han had swollen to several thousand men by the end of the war, most engaged in intercepting. So hungry was it for competent personnel that it did something almost unheard-of in misogynistic Japan: it employed women—putting about 30 nisei girls to work eavesdropping on American radiotelephone conversations. By the middle of 1943 it had outgrown its quarters, and the traffic analysis section moved to the third floor of the Naval War College in Tokyo, leaving only the cryptanalysts at the Navy Ministry

  They comprised the 2nd Branch of the Tokumu Han’s three. In charge was Captain Endo. Under him were several national sections: United States and Britain, with about 50 officers under Lieutenant Commander T. Satake; China, with about 20 officers under Lieutenant Commander Nakatani; Russia under Lieutenant Commander Masayoshi Funoto; and Italian, German, French, and others, about 10 officers. The 3rd Branch handled traffic analysis. It was likewise organized on a national basis, subdivided into areas, with an average of two officers and a handful of enlisted men working on each area. This was fluid, however, and sometimes as many as ten officers would be working on a single area. The branch was commanded by Morikawa, now a captain, who, in a separate capacity, also headed the Owada Communications Unit. The 1st Branch planned, made policy, and distributed the results of t
he two operating branches. In charge was Captain Amano, with Commander Hideo Ozawa his executive officer. Command of the entire Tokumu Han was vested in the chief of its parent body, the 4th Department; in effect, this gave the Tokumu Han a seat on the Naval General Staff. In 1943 the head of the 4th Department was Rear Admiral Gonichiro Kakimoto, and at the end of the war, Rear Admiral Tomekichi Nomura.

  The Imperial Japanese Navy radio intelligence organization (Tokumu Han) of the 4th Department (communications), Naval General Staff

  In sharp distinction to American cryptanalysts, who were reading the vast majority of Japanese messages, including those in the cryptosystems of topmost security, the codebreakers of the Tokumu Han failed almost completely in extracting usable information from American messages. They did not even attempt to solve medium- and high-echelon messages, couched in cryptosystems far beyond their ability. They concentrated instead on three simpler cryptosystems of the lowest levels of command. Even with these, they achieved only limited success.

  Typical was their experience with a small code that they called AN 103. Carried by U.S. Navy patrol planes, it consisted of a few dozen expressions, such as enemy sighted. The code was changed every seven to ten days, but the same plaintext expressions appeared in successive editions, facilitating solution. Fortunately, such solutions were usually obtained too late to take any action based on them.

  The Tokumu Han cryptanalysts succeeded best with BAMS, the two-part superenciphered Allied merchant ship code. They solved about half of the BAMS intercepts. How were they suddenly able to do so well with so relatively difficult a system? Germany had given them the basic BAMS codebook, which had been captured by her raider Atlantis. Consequently, the Japanese had only to remove the superencipherment. BAMS provided occasional tidbits of information—three transports had departed from California, for example, or a vessel’s course and speed data—but even here, Ozawa complained, “By the time the code [message] was broken, the ship was no longer in the original area.”

  The Tokumu Han expended most of its cryptanalytical energies on the CSP 642, the strip cipher, which the U.S. Navy regarded as its lowest-echelon system. The Navy complicated it by not using the full complement of 30 strips every time. Instead it eliminated from zero to five strips from one day to another. Thus one day’s messages might use only 25 strips, the next day’s, 27, the next, 30.

  Japan had captured strip ciphers on Wake and Kiska, and with these she attacked the intercepts. Her methods mixed sophistication and naivete. To determine how many strips had been eliminated, the Tokumu Han used I.B.M. tabulators of the First Life and the Meiji Life Insurance companies of Tokyo. These took frequency counts at intervals of 30, 29, 28, …, 25 and compared them; the interval that showed the most repetitions indicated the correct encipherment length. Many of the strip messages were sent by American submarines; these were identifiable by their indicators—BIMEC or FEMYH—and by their transmission from close to the Japanese coast. The Tokumu Han could know that at that position a merchant ship had been sunk, or that certain units of the Japanese fleet near there were steaming at such-and-such a course and speed, and that the submarine was reporting this. With this as a lead, two first lieutenants who had majored in English in college, Shimizu and Oda, composed what they thought the plaintext intercept was. They varied expressions, word positions, guesses of latitude and longitude until they had a supposed plaintext that matched the cryptogram in length and whose letters all differed from their ciphertexts—since in the strip system no letter can represent itself. Then they arranged and rearranged the strips until they had reproduced the ciphertext on one line and the presumed plaintext on another; the sequence of strips almost certainly represented that day’s key. With it they decrypted other intercepts.

  This tortuous method—for some reason they did not heed the lessons of de Viaris and Friedman on solving this system—suggests why so little information was extracted from the strip cipher. The Tokumu Han kept increasing the size of the section in its American branch that handled strip messages until there were about 40 officers, 10 enlisted men, a dozen typists, two dozen women clerks, Professor Yamanashi of the Navy War College, and a mathematician, Ozaki. Though efforts were continued up to the end of the war, the life had long since gone out of them; the Tokumu Han, considering the strip cipher unbreakable for all intents and purposes, vacated its hopes for cryptanalysis and looked instead to traffic analysis as its chief source of information.

  The difficulty with this, as Lieutenant Commander Satake put it, was that “Our whole analysis was based on probabilities; there was nothing of a definite nature.” The 3rd Branch graphed the volumes of urgent, priority, routine, and deferred messages transmitted from each major American station. It charted the traffic flow among the various call-signs. It located the transmitters by a widespread direction-finder net of a dozen linked stations situated from Kiska to Rabaul, from Wake to Manila. By following the bulge in BAMS transmissions from California to Hawaii to, say, Guam, the traffic analysts could predict the general area in which the next American assault would come. Messages from reconnoitering submarines or airplanes reinforced the estimate. The time of the attack was often gauged by noncommunications means—such as guesses based on previous movements—but sometimes by such communications intelligence as the imposition of radio silence or an increase in the urgency of reconnaissance messages. None of these methods, however, enabled the 3rd Branch to pinpoint time or place. The Japanese knew in advance, for example, that the United States was mounting an invasion of the Philippines, but when it would come they could tell no more closely than within a month, and upon which island the assault would fall, they never knew until it happened. Compared to the crystalline precision of America’s Midway intelligence, Japanese intelligence floundered in a miasma of vaporous generalities. Only once in four years of war—at the Marshalls—did it get word to a garrison early enough to help it prepare for an impending attack.

  The Japanese Army, personified by the combined War and Prime Minister General Hideki Tojo, had panted for this war much more than the Navy, and so might have been expected to produce striking communications-intelligence results when the desired hostilities broke out. The woeful actuality was summed up in one sentence after the defeat of Nippon by Lieutenant General Seizo Arisue, chief of Army intelligence: “We couldn’t break your codes at all.”

  It was not for lack of trying. The Army centered its communications-intelligence work at Tanashi and dispersed seven intercept and direction-finding units through the home islands alone. Runners, telegraph, and radio brought the average of 250 diplomatic and press messages and 800 military dispatches intercepted each day to the cryptanalysts. Headed by Major Machida, they worked in an old folks home named Yofuen in Tokyo, with two unimportant subordinate groups in the villages of Ono and Itakura. They failed utterly with the diplomatic traffic. Not until 1944 did they begin work on the military strip ciphers, and though they drafted some mathematics students, brought in an I.B.M. tabulator, and consulted with the Navy, they had no more luck than did the Tokumu Han.

  Field units were attached to Army staffs. They listened in to American radio messages and even sent out special wiretapping patrols. Results were disappointing, mainly because few men at the front could understand English. And since the soldiers’ war in the Pacific comprised a series of brief, individual battles for small islands, the Army had little opportunity to build up an enemy order of battle or to predict attacks by traffic analysis. As for field cryptanalysis of Allied messages, Arisue’s mournful plaint was echoed by a colonel on the staff of the 25th Army: “We did not break your codes.” How debilitated their intelligence must have been—for the Japanese rated communications intelligence as their most valuable source of information on the enemy!

  The Arisue concession must be qualified somewhat. In occupied Manila, cryptanalysts in a staff section monitoring squad of the 14th Army frequently solved the messages of Filipino and American guerrillas. These jungle warriors wirelessed in
formation on Japanese activities to MacArthur’s headquarters in Australia and direct to San Leandro, California. At first they enciphered with whatever was at hand. One of the first guerrillas, an unsurrendered American soldier on Luzon, employed the venerable U.S. Army Cipher Disk. Some used M-94s, at least one of which was captured by the Japanese. When Colonel Wendell Fertig on Mindanao finally contacted Australia early in 1943, he was instructed: “If you know double transposition use as key first name of second next of kin [which would be his first-born child, Patricia] and city of residence second next of kin [she was then living in Golden, Colorado] and encode the following information …” as an authentication. Later in the war, the guerrillas employed new ciphers smuggled in on the supply submarines. One such system, comprising seven closely typed pages and intended for a special operation, was microfilmed and concealed in the ankle patch of a pair of sneakers for transportation to another leader, Macario Peralta, Jr., on Panay.

  The cryptanalysts in Manila seemed to do best in the first half of 1943—after volume from the guerrilla stations had built up but before the improved systems appeared. From February through April they read messages in a number cipher emanating from Cebu and used to report shipping movements, though they failed to solve “a special code … used by one part of the Cebu system”—probably the guerrilla units under Colonel Harry Fenton. In March and April they broke the system used by Peralta’s units on Negros, as well as various double transpositions, until their keywords were changed in April. Solution of the system used by the headquarters station DKZ provided information “on the general organization of the enemy guerrillas over all of Negros, Siquijor, and Mindanao,” boasted the cryptanalysts in their report for the last ten days of April. Next month they had to confess that “deciphering is at a standstill,” but in July—perhaps as a result of an increase of traffic, since 214 messages were sent in just ten days by only two stations, KML and WZE—they broke through to their last big success. They cracked the Fenton system that had rebuffed them before, as well as Fertig’s messages, enabling them to read both leaders’ back files since March. A few days later, interrogation of a captured American yielded keywords used for communication with Australia and America.