The Polyalphabetic Revolution
"Le chiffre indéchiffrable" — The unbreakable cipher. Until it wasn't.
Polyalphabetic systems use changing substitution alphabets. They obscure simple letter-frequency relationships, but repeating keys create structure that analysts can exploit. Bellaso, Vigenère, Babbage, Kasiski and Friedman contributed different constructions or methods; the history is more varied than a single cipher remaining universally trusted for 300 years.
The key insight: A key selects which alphabet applies at each position. The same plaintext letter can therefore map to different ciphertext letters. Repeated key positions may still be separated into simpler substitution problems.
The modern repeating-key cipher bears Vigenère’s name; Bellaso’s 1553 publication belongs to its earlier history. Babbage and Kasiski developed attacks in the nineteenth century.
Key: LEMONLEMONLE
Cipher: LXFOPVEFRNHR
Beaufort cipher (1857) — Sir Francis Beaufort's reciprocal Vigenère variant where encrypt and decrypt are the same operation. Broken by Kasiski examination.
C = (Key − Plain) mod 26
Porta cipher (1563) — della Porta's reciprocal polyalphabetic table where each key letter swaps two halves of the alphabet. Falls to Kasiski just like Vigenère.
Reciprocal: same operation decrypts
A numeric-key Vigenère variant traditionally associated with Gronsfeld in the seventeenth century. Ten shifts per position reduce the key choices.
Shift by digit value — only 10 options per position
Running Key cipher (19th c.) — a Vigenère variant using a long passage of natural text as the key. Broken by Friedman with high-probability-trigram dragging.
Key: FROMTHISBOOK
Plaintext autokey extends a primer with message letters. It removes a fixed repeating keyword but retains exploitable language structure and error propagation.
Key: KEYATTACKATD
How the Repeating Key Betrays Itself
The same plaintext sequence under the same repeating-key alignment produces the same ciphertext. Repeated ciphertext sequences can therefore suggest distances divisible by the period. Coincidental repeats also occur, so candidate periods need testing. The repeated string below is an illustration of the inference, not a guarantee that its key length is known.
Ciphertext:
LXFOPVEFRNHRLXFOPVEFRNHR
↑ ↑
pos 6 pos 18
Distance = 12 → test divisors of 12; coincidence is also possible
The index of coincidence measures the chance that two positions chosen without replacement contain the same letter. English samples often lie near 0.066, while independent uniform A–Z samples approach 1/26. Repeating-key substitution can move the overall score toward the latter, but length, language and key period matter: IC supplies evidence, not a certain cipher identification.
| Text Type | IC Value |
|---|---|
| Random | ~0.038 |
| Illustrative Vigenère sample | ~0.052 |
| English | ~0.066 |
Vigenère repeats a short sequence of letter shifts. ChaCha20 instead generates pseudorandom bit blocks from a key, nonce and counter, then combines them with plaintext using XOR. Its security is computational, and its usage rules prevent keystream reuse. “Non-repeating” does not mean that individual output blocks can never coincide.
The running key cipher is one step from the one-time pad. The difference: a book passage is not random. If the source text has linguistic structure, that structure leaks into the ciphertext and enables statistical attacks.
The Vigenère lesson: short repeating keys are fatal. Modern systems can reuse a suitably strong key for multiple messages, but must respect each scheme’s nonce, counter, and usage limits. Key derivation can separate session keys; it does not automatically guarantee unique key material.