Purple

History

The Japanese Foreign Ministry’s 97-shiki ōbun injiki carried its diplomatic traffic from 1939. The U.S. Army’s Signal Intelligence Service broke it in 1940 without ever seeing one, from messages in depth and known plaintext, and built an analog; Friedman’s report of October 1940 tells how. The Navy’s manual for its own copy, R.I.P. 77, prints the switches’ wiring. The fourteen parts of Japan’s note of 7 December 1941 were among the messages read in Washington; their decrypts are in the National Archives. No Japanese machine survives whole.

How it works

A plugboard sends six letters, set by the day’s alphabet, through the sixes switch and twenty through three twenties switches, each a stepping switch of 25 positions; the same plugging takes the output back to letters. The sixes switch steps every letter and one twenties switch with it: the fast, except at the sixes’ 25, when the middle steps, or at 24 with the middle at 25, when the slow does. The key gives the sixes’ start, the three switches’ starts and which is fast and which middle.

Without the key, the output plugboard only relabels letters, so a setting is scored letter by letter, each ciphertext letter by the lead whose outputs coincide most, with no plugboard to search. First the sixes’ start and letters, then all 93,750 settings of the twenties; for the best, the leads are climbed, a slow switch the coincidences cannot see tried at its 25 positions, and the alphabet polished by the language’s quadgrams.

How it was checked

Every entry of the four switch tables agrees with R.I.P. 77’s pages; R.I.P. 77’s two tables of the switches’ positions and the paper’s figures come out step for step. Part 1 of the 14-part message reads its first 216 letters exactly as Freeman, Sullivan and Weierud print them, and the Tokyo–Berlin message of 27 July 1942 shows their markers at group 41. From the ciphertext alone the search finds part 1’s and part 5’s keys and alphabets exactly.

Limits

The tables are the analog’s, in the SIS’s numbering of the switches; the Japanese numbered them the other way about. Japanese plaintext, in romaji, has no statistics here, so the search reads English and the other languages of the pills; a Japanese message is read with its key. Under 600 letters the right setting does not stand out; under 1,000, not always. The key books’ indicators and monthly additives are not offered.