Nomenclator

IPhenomenon

A Venetian nomenclator of the sixteenth century has three parts: an alphabet of two- or three-digit codes for letters, a list of syllables for the sounds its language needs, and a nomenclatura of hundreds of proper nouns each with its own number.

The bench beside this entry reproduces the shape of it. Letters get two- and three-digit codes; THE, AND, KING, POPE, WAR and FRANCE get codes of their own; words are separated with a slash.

The last row is the real craft. If E has four different codes and the encipherer rotates between them, the frequency profile flattens out and the classical attack weakens. But the attacker no longer needs the letters anyway, because the names are the message: a dispatch consisting entirely of 03 06 05 needs no cryptanalysis at all if you have captured a copy of the book.

IIKey space

The key space is the number of entries in the book, and it is entirely administered rather than computed. A diplomatic nomenclator might carry 1,200 codes; a papal one, more.

The advantage is real: the book does something no algorithm can, which is to make the important words the expensive words. Frequency analysis assumes every plaintext symbol costs the attacker the same amount of work. A nomenclator deliberately breaks that assumption.

The cost is equally real, and it is operational: every correspondent needs their own book, every book must be delivered securely, every book must be replaced when it might have been captured, and every clerk must be able to use it accurately under time pressure.

The nomenclator's security parameter is not entropy. It is the number of people who must keep a secret.

IIIEnciphering

• Compile the book: letter codes, common-word codes, and a proper-noun list. Number the entries in a dense range.
• Add homophones: give the most frequent letters several codes each, and add unused numbers that mean nothing.
• Encipher each word: use the word code if one exists, otherwise encipher letter by letter.
• Rotate homophones by a rule the receiver knows — strictly in order is the simplest and the most common in surviving books.
• Separate words clearly, usually with a slash or a doubled code. Ambiguity between the letter layer and the word layer is the most frequent source of error.
• Insert nulls at the end of sentences so that message length carries no information.
• Destroy the working copy of the message. The book is the only artefact that should survive.

IVDeciphering

• Identify word boundaries first, from the separators.
• Look each code up in the book. Two- and three-digit codes are letters; four-digit codes starting with a distinctive range are usually proper nouns.
• Expand homophones: if the same letter has four codes, all four map to the same letter, so concordance is straightforward.
• If a code has no entry, the book is either out of date or from a different correspondent — this is the single most common practical failure.
• Fill in the nulls as blanks and read the dispatch.
• Remember what you are actually holding: with the book, this is not cryptanalysis, it is transcription. The security was administrative, so the break must be too.

VWeakness

The book is a physical object. It can be captured, sold, copied, or photographed. Mary, Queen of Scots was undone not by cryptanalysis of her nomenclator but by the seizure of the decoded plaintext of her own letters — the cipher held, the operational security did not.

Repetition inside the code layer. Even with homophones, the same code appears for the same word, and if the enemy knows one name, they know it every time it appears.

Book maintenance scales badly. Adding a new general to a campaign means reissuing the book to every correspondent, and reissuing is slow, expensive and visible.

Compromise is total and silent. A broken algorithm fails for everyone at once and is noticed. A captured codebook fails for one correspondent and might not be noticed for months, during which every message is read.

Clerical error is common. Codes are four digits in a dense column; mis-transcription produces a plausible-looking ciphertext that decrypts to nonsense, and the standard remedy — resend — doubles the traffic available to the attacker.

VIPractice

·History note

The term comes from the Latin nomenclator, the slave whose job in a Roman household was to whisper the names of approaching citizens to his master. The cryptographic nomenclator emerged in the Italian city-states in the fourteenth and fifteenth centuries and was the mainstay of Venetian, papal, Spanish and English diplomatic correspondence. The Babington Plot letters of 1586, in which Mary, Queen of Scots was implicated, were enciphered with a nomenclator of roughly 200 symbols mixed with letter codes; the English secretariat, working with Thomas Phelippes, read them and, decisively, intercepted the sealed postscript that authorised the plot.

·Further reading

The nomenclator's modern descendant is the codebook cipher of the twentieth century and, indirectly, the one-time pad: the recognition that what you need is not a cleverer transformation but key material that is used once and then destroyed.

The other descendant is the idea of protecting selected content differently from the rest. Modern systems do this too, with different keys and different lifetimes for different classes of data — an ordinary message and a signing key are not protected to the same standard, and should not be.

Next on the loom: null-cipher hides the message by not looking like a message at all.

+Weave into