Unveil

Code

Reading a text without its key: naming the cipher, peeling off encodings, counting letters, measuring an unknown script, breaking nomenclators, reading messages in depth, replaying the Enigma breaks, checking anagrams.

Unknown Script

Measures a transcription of signs that no one reads: each sign’s place in the word and the line, which follows which, the words by rank (Zipf), Shannon’s F1 and F2, and Sukhotin’s split into vowels and consonants (1962), as Guy put it into English in 1991 — each beside the same signs shuffled or six languages drawn at the text’s length. It measures; it reads nothing.

Terms

History, how it works, how it was checked, limits

History

Before a script or a cipher of signs is read, it can be counted. Shannon defined in 1951 F_N, the entropy of a letter given the letters before it. B. V. Sukhotin published in Russian in 1962 an algorithm that splits an alphabet into vowels and consonants by contacts alone; Jacques Guy gave it in English in Cryptologia in 1991, for simple substitution ciphers. The Borg cipher’s decipherers found Latin the closest of sixteen languages before reading it (2017).

How it works

A sign is one character, or a code between spaces; word dividers and signs left out are named. Each sign is counted at its places in the word and the line, and each pair side by side, against the count its share gives: a cell is marked only where chance gives as many, or as few, less than once in a hundred tables with every cell counted. F1, F2 and Zipf’s plot are set beside the signs shuffled and the six languages drawn at the text’s length, F2 also under a Vigenère key of five. Sukhotin’s algorithm is stepped as Guy writes it, its split set beside the split found in the signs shuffled. The shape of the shares and the index of coincidence are set against each language drawn at the text’s length, and a text in Latin letters is read by each language’s statistics.

How it was checked

Guy’s own example, SAGITTA round a plate, gives his matrix and his sums step by step, A and I its vowels. On the Borg cipher the first five signs taken for vowels are those its key reads a, e, i, o and u, and its signs fit Latin nearest; on Poe’s cryptogram the algorithm takes t for a vowel, as Guy says it does in English. The Copiale’s signs fit no language. The languages drawn give F1 and F2 as counted over their books. The figures are held to the counts.

The tests hold 56 checks of this tool; 8 rest on outside material:

  • A historical message The Borg cipher (Aldarrab, Knight and Megyesi, 2017): the first five signs Sukhotin’s algorithm takes for vowels, w x 6 o d, are those the decipherers’ key reads e, i, a, o and u; then 4 (f) and E
  • A historical message The Borg cipher’s signs fit Latin nearest of the six, as Aldarrab, Knight and Megyesi (2017) found Latin the closest of the languages they tried; Italian and Spanish fit too; it is in no Latin letters, so fluency is not read
  • A historical message The Borg cipher’s folio as Örneholm read it (2018) is read as Latin: fluency 94 of 100, the next French at 72, its letters shuffled at most 20 in Latin
  • A print’s own example Guy’s Step 1 (Cryptologia 15:3, 1991, p. 259): SAGITTA written round a plate gives his symmetrical matrix of contacts over S A G I T, TT 2 on its diagonal
  • A print’s own example Guy’s Steps 2–5 (pp. 259–260): the diagonal zeroed, the sums 2 4 2 2 2; A a vowel, the sums −2 4 0 2 0; then I, −2 4 −2 2 −2, and no consonant left above zero
  • A print’s own example Guy’s result (p. 260): A and I are the vowels of SAGITTA, S, G and T its consonants
  • A print’s own example Poe’s Gold-Bug cryptogram (1843): Sukhotin’s algorithm takes 8 ; 6 ‡ 5 for vowels first — Poe’s e, t, i, o and a: the t for TH, as Guy’s note 3 (p. 261) says it does on English — then 3 and —
  • A print’s own example Guy’s table at his last step (p. 260): each row the matrix with its diagonal zeroed and its sum, S −2, A 4, G −2, I 2, T −2; I the vowel found now

Every tool’s checks

Limits

It measures and reads nothing, and holds no corpus. A short text says little, and the mark of beyond chance is strict: a rare sign’s habit goes unmarked. Chance takes each sign as independent of its neighbours. Sukhotin’s algorithm assumes an alphabet and errs on a short text: on the Borg folio’s Latin it takes f for a vowel. Only six languages, a simple substitution and a Vigenère key are drawn; no homophonic, syllabic or verbose cipher. Sukhotin’s Russian was not read.