Hill Cipher

Hill Cipher is a polyalphabetic cipher created by extending the Affine cipher, using linear algebra and modular arithmetic via a numeric matrix that serves as an encryption and decryption key.

Hill cipher encryption uses an alphabet and a square matrix $ M $ of size $ n $ made up of integers numbers and called encryption matrix.

Split the text into $ n $-grams. Complete any final incomplete ngrams with random letters if necessary.

Substitute the letters of the plain message by a value: their rank in the alphabet starting from $ 0 $.

For each group of values $ P $ of the plain text (mathematically equivalent to a vector of size $ n $), compute the multiplication">matrix product: $$ M.P \equiv C \mod 26 $$ where $ C $ is the calculated vector (a group) of ciphered values and $ 26 $ the alphabet length.

From cipher values $ C $, retrieve cipher letters of the same rank in the alphabet.

Hill cipher decryption needs the matrix and the alphabet used. Decryption involves matrix computations such as matrix inversion, and arithmetic calculations such as modular inverse.

To decrypt hill ciphertext, compute the matrix inverse modulo 26 (where 26 is the alphabet length), requiring the matrix to be invertible.

Decryption consists in encrypting the ciphertext with the inverse matrix.

The determinant of the Hill matrix must be prime with 26 to ensure that the matrix is invertible modulo 26 (the value 26 comes from the length of the Latin alphabet having 26 letters).

This ensures that each modular multiplication operation remains one-to-one (each input vector/ngram corresponds to a single encrypted vector/ngram and vice versa).

If this were not the case, several different vectors/ngrams could be encrypted in the same way, which would make decryption ambiguous and difficult to perform reliably.

For a 2x2 matrix, the 4 numbers $ \{ a,b,c,d \} $ must satisfy the condition that $ ad-bc $ is coprime with 26.

The ciphered message has a small index of coincidence and similar ngrams can be coded using the same letters.

Any reference to an actual hill or mountain is a clue.

Sometimes the groups of letters are left visible (all of length n = 2, 3 or 4) which suggests that the matrix is of size n.

dCode proposes to bruteforce test around 6000 combinations of 2x2 matrices (with digits between 1 and 9) and alphabets.

For matrices containing numbers $ >= 10 $ or larger size matrices, computation times become exponentially longer.

Hill is already a variant of Affine cipher. Few variants, except the use of large size matrices.

Hill cipher has been created in 1929 by Lester S. Hill