BAYT AL-ḤIKMA · A TREATISE ON EXTRACTING THE OBSCURED
The cipher betrayed by its own letters.
Around 850, by most accounts, the philosopher Abū Yūsuf al-Kindī set down the oldest surviving description of frequency analysis: count the signs of a hidden text, set the tally against the habits of ordinary speech, and the mask comes off. Below, an intercepted dispatch in a substitution cipher is being counted apart in real time. Nothing is staged — the count is the crack.
THE COUNT
one alphabet traded for another · gold is still cipherfaint not yet counted · gold counted, unread · ivory fixed · pale slant tentative
counting…
THE COMPARISON
Counted signs (gold, above the rule) against the letter frequencies of printed English (verdigris, below), both ranked. Reference after Lewand, Cryptological Mathematics (2000). When the two spines share a shape, the cipher is done for.
THE FIXING
recovered keyA sign is fixed when its reading has held steady through five consecutive rounds of refinement with at least eight examples counted. A fixing that later fails the word test is revised — marked in verdigris.
THE METHOD
Al-Kindī's procedure asks for no key, no traitor, no seizure of the courier — only patience and arithmetic. Three movements, exactly as the instrument above performs them:
Count every sign
Tally each symbol of the hidden text, one by one, as it arrives. The histogram is built from the ciphertext itself — every bar is a measured count, never an illustration.
Set tally against tongue
Rank the counted signs beside the ranked letters of ordinary writing — e, t, a, o, i, n… The most frequent sign confesses the most common letter; the shape of one spine betrays the other.
Fix the frequent first
Lock the confident readings and let each fixed letter test its neighbours in the words it enters, revising the middle ranks until the passage stands clear. The rarest signs resist longest — with one or two examples, statistics has nothing to grip.
TOO FEW SIGNS
Frequency analysis needs material. Drag the intercept down and the method visibly starves: below a few hundred signs the columns of the count wobble, the middle ranks scramble, and the tentative reading collapses into garble. A few dozen signs will crown the wrong king. That limit is not a flaw of the instrument — it is the mathematics. A rare letter seen twice tells you almost nothing; a common one must be seen dozens of times before its rank holds steady from one page to the next.
Try it against the full dispatch of 1,304 signs:
THE MANUSCRIPT
Abū Yūsuf Yaʿqūb ibn Isḥāq al-Kindī — philosopher of the Abbasid court, tutor in the House of Wisdom, author on optics, music, medicine and arithmetic — wrote his Risāla fī istikhrāj al-muʿammā, a treatise on extracting obscured correspondence, in Baghdad in the ninth century. It described letter frequency, probable words, and vowel habits as weapons against the substitution ciphers of the chancelleries. The treatise survived in a single known copy, found again in Istanbul's Süleymaniye collections in the 1980s — roughly six centuries before Europe's cipher secretaries wrote the same method down.
One page of counting undoes a century of locksmithing: the lock was made of letters, and letters keep their old frequencies even in disguise. THE DISPATCH ABOVE, ITSELF — PARAPHRASED
Honesty of the demonstration: al-Kindī counted Arabic; this page demonstrates his method on English — a different alphabet with the same disease. The comparison spine is Lewand's published table; the word-habit scoring that refines the middle ranks is counted at load from a reference folio of 2,087 letters of ordinary English embedded in this page. Every number in the readout is measured from the running text.