War with JapanConverted for the Web from "Battle Of Wits: The Complete Story Of Codebreaking In World War II" by Stephen Budiansky
Jump to: Doolittle's Raiders | Admiral Isoroku Yamamoto
Joseph J. Rochefort | Stealing the Japanese Code | War Against Japan
After Pearl Harbor | A Man with a Mission | War Against Bureaucracy
Ambush the Ambushers | The Battle of Midway
Denouement: Intelligence Vindicated
America's very success, in September 1940, in breaking the Japanese diplomatic cipher, code named "Purple," had the ironic effect of distracting attention from where it could have been more profitably focused in the fateful months leading up to Pearl Harbor. The Purple cipher carried the highest-level diplomatic messages of the Japanese Empire; this was intelligence of such remarkable value that it was given the code name magic. The Purple cipher was generated by a complex machine. It used a cascade of rotating switches to encipher every letter of a message in a different key from the last or the next. In one position of the switches the letter A would become G; in the next it would become P. The U.S. Army's code breakers had, in eighteen months of intense effort, deduced the wiring and setup of the machine without ever seeing one, a feat of pure analysis the likes of which had scarcely before been seen. After hastily soldering together telephone switches and relays to produce a replica of the machine, they proceeded to decode the Japanese messages almost as quickly as they arrived.
On the morning of December 3, 1941, a Purple message came through ordering Japan's embassy in Washington to destroy its code books, and even one of its two vital Purple machines. Frank Rowlett, a senior cryptanalyst of the Army's Signal Intelligence Service, arrived at his office at noon that day from a meeting, plucked this latest magic decrypt from his in-box, and proceeded to read its contents with mounting incredulity. With only a single machine it would obviously be impossible for the embassy to continue its normal flow of business. Colonel Otis Sadtler, who was in charge of distributing the magic decrypts, showed up in Rowlett's office at that moment and began to pepper him with questions. Had the Japanese ever sent anything like this before? Could they be getting ready to change their codes? Perhaps they suspected their current codes had been broken? Then the only possible meaning of this extraordinary message sank in. Sadtler pulled himself to attention. "Rowlett, do you know what this means? It means Japan is about to go to war with the United States!" And, decrypt in hand, Sadtler took off literally running down the corridor of the Munitions Building to alert the head of Army intelligence.
On the night of December 6, an aide interrupted the President at a White House dinner to deliver him the latest magic decrypts. These erased all remaining doubt. Japan was preparing to break off diplomatic relations. War was inevitable.
But diplomatic communications are not the place where military orders are delivered. America knew that Japan was going to strike; it did not know where she would strike. To know that would require breaking into the Japanese naval codes, and there was only one catch: Since mid-1939, America had not read a single message in the main Japanese naval code on the same day it had been sent. For most of the period from June 1, 1939, to December 7, 1941, the Navy was working on naval messages that were months, or even over a year, old.
Partly this was a matter of manpower, partly it was a matter of human nature. Magic was such a dazzling prize that it blinded its possessors to the smaller but sometimes more valuable gems that lay buried among the dross and slag of supply orders and fleet maneuvers. JN-25 was the most recent descendant of the Japanese Navy's Red code; like its predecessors it was an enciphered code. At the time the new code first appeared on June 1, 1939, the U.S. Navy's Washington code breaking staff had grown to about thirty-six hands. By this time the "research desk" had acquired the official bureaucratic designation of OP-20-G, designating it as part of the Office of Naval Communications, OP-20, and Safford was back in charge after several tours of sea duty. The staff of thirty-six included translators, clerks, radio direction-finding experts, intelligence analysts, and officers responsible for the security of the Navy's own codes; only a handful were trained cryptanalysts, and of these only two or three could be spared to work on the new code, which was initially given the designation AN-1.
Over the course of months they laboriously punched every message on IBM cards and searched for any clues that would give them a toehold on this completely uncharted terrain. But everything about it had the air of scholarly inquiry, far from the heat or urgency of battle. AN-1 was a "research" project, not a "current decryption" job; reconstructing the meaning of thirty thousand code groups and piecing together thirty thousand random additives was not going to be the work of a moment. The Japanese Navy's radio signals were intercepted by U.S. Navy operators in Hawaii, Guam, and the Philippines. The operators transcribed the Morse code signals by hand onto message blanks, bundled them up, and once a week handed them over to the captain of one of the Dollar Line's "President" passenger liners that plied the Pacific. The captains, all members of the Naval Reserve and therefore authorized to act as couriers of confidential documents, dropped the packages in the mail to Washington when their ships reached the West Coast. A very small amount of priority traffic could be entrusted to the Pacific "Clippers" of Pan American Airways; in the hull of each of these airplanes a small steel strongbox had been welded into place just for this purpose, the keys held by naval officers along the route. But delays of weeks from the time a message was transmitted by the Japanese to the time it arrived in Washington were the norm.
Throughout 1939, 1940, and 1941 a slow-motion cat and mouse game ensued. Sorting the IBM cards by number and printing out huge catalogues of every code group in every message, the Navy cryptanalysts began to detect a few ghosts of the underlying code. The numbers in each message that appeared to serve as indicators, telling the recipient what page of the additive book had been used, were not quite as perfectly random as they should have been. The numbers bunched up, meaning that some lazy Japanese code clerks were reusing the same pages of additive over and over. That was a classic error; messages enciphered with the same additive pages could in principle be cracked. It was not until fall of 1940, however, that the first real break came -- and it came just in time to be rendered obsolete by a completely new, and much more complex, code book that the Japanese brought into service on December 1, 1940.
By the summer of 1941, as tensions in the Pacific grew, every section of OP-20-G was desperately short of help. Messages in the Purple cipher, which could be read in their totality and almost always the same day they were transmitted, claimed first priority. Meanwhile seven thousand AN-1 messages were pouring in each month by mail to Washington, while only sixteen men could be spared to work on them. As the Navy began calling up Naval Reserve officers throughout the summer and fall, that figure crept up by about one per month. Station Cast, the Navy's intercept station at Cavite in the Philippines, was also trying its hand at compiling the reams of printouts and worksheets needed to tease apart the AN code and its additives; since spring 1941 a highly secret collaboration between Cavite and the British government's code breakers in Singapore had been under way on the project as well. But it simply wasn't enough.
Slowly and laboriously, the new code book was being reconstructed; again, inexorably, on August 1, 1941, the Japanese introduced a new, 50,000-group additive book that sent the code breakers back to the beginning. By November 1941 only 3,800 code groups had been identified, along with only 2,500 additives reconstructed in the current system. It was far less than 10 percent of the total, nowhere near enough to read current traffic.
Conspiracy theorists continue to weave elaborate scenarios "proving" that America had advance warning of the Japanese attack, with one branch of the "FDR knew" theorizers insisting that AN traffic was in fact being read in 1941. Yet month-by-month progress reports, internal histories, war diaries, logs -- some declassifed only in 1998 -- are all in agreement: Not a single AN message had ever been read currently by the time of Pearl Harbor, and not a single AN message transmitted at any time during 1941 was read by December 7.
Five years later, with the war safely won, a few of OP-20-G's cryptanalysts were tidying up loose ends and decided to go back and try to crack the unread AN-1 traffic that had piled up in the months just before Pearl Harbor. What they found was enough to break an intelligence officer's heart. Over and over, the orders to the Japanese fleet during October and November 1941 repeated a single theme: Complete all preparations and be on a total war footing by November 20. Several messages referred to exercises in "ambushing" the "U.S. enemy." And one signal, dispatched November 4, ordered a destroyer to pick up torpedoes that Carrier Divisions 1 and 2 "are to fire against anchored capital ships on the morning in question." None specifically mentioned Pearl Harbor, and indeed many other intelligence indications in those critical months pointed to the Philippines, or even the Panama Canal, as possible targets of Japanese naval action if war broke out. Yet the prePearl Harbor AN traffic, had it been broken at the time, would certainly have conveyed heavy hints of what was to come.
Jump to: Doolittle's Raiders |
Admiral Isoroku Yamamoto
Copyright © 2000 by Stephen Budiansky. All rights reserved. Converted for the Web with the permission of Simon & Schuster.
Click to Amazon to buy "Battle Of Wits: The Complete Story Of Codebreaking In World War II."