Mary Queen of Scots — The Castelnau Letters

Mary's Castelnau cipher is a sophisticated Renaissance nomenclator — the same class of system used in the Babington Plot correspondence (1586), which was broken by Thomas Phelippes. The Castelnau cipher is substantially more complex:

• Symbol alphabet: ~100 purpose-designed symbols representing letters and common syllables, not the Roman alphabet, making frequency-alphabet mapping non-trivial.
• Code groups: Named persons, places, and frequent political terms replaced by dedicated codewords (e.g., a specific symbol for "Monsieur de Castelnau," another for "Her Majesty the Queen of England").
• Nulls: Non-semantic symbols interspersed to inflate symbol frequency counts and frustrate statistical attacks.
• Superencipherment layer: Some passages appear double-encoded — encoding already-encoded text — as an additional security measure for the most sensitive passages.

The plaintext language is French — an additional barrier for analysts who might expect English or Latin in Mary's correspondence.

The attack method, developed by George Lasry, is a variant of the simulated annealing / hill-climbing approach used in modern historical cryptanalysis. The key components:

1. Transcription: Lasry manually transcribed all ~50,000 cipher characters into a machine-readable format, assigning each symbol a numeric ID. This step took weeks.
2. Null identification: Statistical analysis of symbol frequency distributions identified probable null characters (symbols appearing with anomalous frequency patterns inconsistent with natural French).
3. Codeword separation: Compound symbols and high-frequency sequences were tentatively identified as code groups (names and words), isolated, and set aside while the substitution alphabet was attacked first.
4. Hill-climbing on n-gram fitness: The substitution key (mapping of cipher symbols to French letter/syllable hypotheses) was optimised using a hill-climbing search scored by French 4-gram and 5-gram text fitness — the probability that a trial decryption matches French natural-language statistics.
5. Manual refinement: Once the hill-climber converged on a near-correct key, Lasry, Megyesi, and Tomokiyo examined readable passages, identified remaining errors, and resolved codeword assignments from context.

The approach is a modern instance of a decades-old technique in historical cryptanalysis, but applied to a cipher of unprecedented size (for hand-written Renaissance matter) and complexity. The full computational work took several months.

Below is a stylised representation of how the nomenclator cipher looks in the original manuscript — numeric cipher symbols where each number represents a letter, syllable, or codeword in the nomenclator:

Each number represents a single item in the nomenclator table. A null symbol (cryptographer's decoy) might appear as · ·98· · or · ·100· · — a high-frequency interspersed symbol that does not contribute to plaintext.

Reconstructed sample entries from the Lasry–Megyesi–Tomokiyo solution (simplified illustration — exact character assignments are in the published paper):

The full content of the deciphered letters is reported in the Cryptologia paper and its supplementary materials. Key categories of newly revealed content include:

• Captivity conditions: Mary describes her treatment at Chartley, her access to correspondence and to her attendants, and her physical health — primary-source details not previously available from the cipher side of the archive.
• Political strategy: Mary outlines her diplomatic positioning — appealing to the French crown for intervention, requesting Castelnau relay messages to specific parties at the French court, and describing her view of her prospects in England.
• Undocumented intermediaries: The letters name individuals not previously known to have been involved in Mary's communications network, expanding the known cast of her diplomatic circle.
• Assessment of Walsingham: Mary's own characterisation of Sir Francis Walsingham and her awareness of surveillance — a rare first-person account of Tudor intelligence from the surveilled rather than the surveiller.

Lasry, G., Megyesi, B., & Tomokiyo, S. (2022). "A ciphertext-only cryptanalysis of a historical nomenclator." Cryptologia, 47(2), 165–195. DOI: 10.1080/01611194.2022.2030632

Original manuscript: Bibliothèque nationale de France, Paris. Département des manuscrits, French MS 2932.

The paper includes the full transcription, the reconstructed symbol/code-group table, the complete plaintext in French, and a commentary on the historical significance. The preprint version was presented at HistoCrypt 2021.

Cryptiana context: Satoshi Tomokiyo's wider Cryptiana work on Renaissance diplomatic ciphers provides the archival and technical backdrop for why the Castelnau nomenclator could be recognised, classified, and finally attacked computationally in the first place.

See also: George Lasry exhibit · Babington Plot Cipher · Further Reading: Cryptologia journal