Reading a text without its key: naming the cipher, peeling off encodings, counting letters, breaking nomenclators, replaying the Enigma breaks of the 1930s and 1940s, checking anagrams.
Nomenclator
Reads and breaks the code tables of the 1400s–1700s: signs for letters, several for the common ones, and for syllables, words and names, with nulls and signs that repeat or cancel. Without the key, letters for the open signs are found by simulated annealing in the chosen language, as Lasry, Biermann and Tomokiyo read Mary Stuart’s letters (2023); rare signs become code words, and nulls are tested one by one.
Terms
About this tool
History
Nomenclators — a cipher alphabet with homophones for the common letters, joined to signs for syllables, words and names — were the ciphers of European courts from the fifteenth century on, and archives still hold letters in them whose keys are lost. Three have been read in the last fifteen years: the Copiale, a German society’s book, by Kevin Knight, Beáta Megyesi and Christiane Schaefer in 2011; the Borg cipher, Latin recipes in the Vatican, by Nada Aldarrab, Knight and Megyesi in 2017; and 57 letters of Mary Stuart, by George Lasry, Norbert Biermann and Satoshi Tomokiyo, published in 2023.
How it works
The ciphertext is written as its transcription and the key as text, a sign a line with what it stands for: letters, a word by its meaning, or a function — a word space, a null, the letter before again, the sign before cancelled. Without the key, simulated annealing gives the open signs letters, scored by the chosen language’s quadgrams less a penalty when the index of coincidence strays from its own. Rare signs and those marked [?] become code words; then, once the reading reads, each sign is tried as a null and kept so if the text reads better without it than under its best letter (in one that does not, every sign would).
How it was checked
The tests read the three exhibits with their published keys: the Copiale’s first page gives the decipherers’ German line for line, folio 8r of the Borg all 234 letters of Örneholm’s Latin, and the 450 signs of Mary’s letter of 20 January 1580 the paper’s text once the edition’s seven changes are made. Without the key, told only the spaces, the doubling colon and the logograms, the solver reads four letters in five of the Copiale’s first page; a French nomenclator of 700 letters made from Louis XIV’s dispatch of 1691, with homophones, nulls and code words, reads 97 per cent, every null found and none false.
Limits
A sign for several letters (the Copiale’s ch, sch, eN) comes out as one letter, not always its own, or as a null, and is set by hand with syllables, words and names; common words with signs of their own are best marked [?]. Under a few hundred letters a homophonic text may not read, and one run in twelve on 1,000 letters of French settles on a false reading, which does not end the others.
- George Lasry, Norbert Biermann and Satoshi Tomokiyo, “Deciphering Mary Stuart’s lost letters from 1578–1584”, Cryptologia 47 (2), 2023, pp. 101–202
- Kevin Knight, Beáta Megyesi and Christiane Schaefer, “The Copiale Cipher”, Proceedings of the 4th Workshop on Building and Using Comparable Corpora (ACL, 2011), pp. 2–9
- Nada Aldarrab, Decipherment of Historical Manuscripts, MS thesis, University of Southern California (2017), §4.2: the Borg cipher
- David Kahn, The Codebreakers: The Story of Secret Writing (New York: Macmillan, 1967)