Banburismus

Banburismus was a probabilistic cryptanalytic method used by the Government Code and Cypher School at Bletchley Park during the Second World War. Developed principally by Alan Turing for the Naval Section known as Hut 8, it extracted evidence about the daily rotor arrangement of the German naval Enigma machine. The method reduced the number of rotor orders requiring examination by a bombe, thereby conserving machine time without directly recovering either plaintext or a complete Enigma key.

The name derived from the large perforated sheets manufactured in Banbury, Oxfordshire, on which intercepted ciphertexts were represented. Banburismus combined physical comparison of these sheets with logarithmic measures of statistical evidence. Its underlying principles included the nonuniform distribution of letters in natural language, the preservation of letter equality by an identical Enigma state, and the mechanical constraints imposed by rotor stepping.

Historical development

German naval communications presented a greater cryptanalytic problem than many contemporary Enigma networks. The Kriegsmarine employed disciplined operating procedures, a larger selection of interchangeable rotors, and message-indicator systems designed to prevent direct reconstruction of the machine setting. For the three-rotor naval Enigma, selecting and ordering three rotors from a set of eight produced 336 possible rotor orders before ring positions, plugboard connections, and initial rotor positions were considered.

Turing developed Banburismus in 1940 as Hut 8 sought a method for ranking these orders before committing scarce bombe capacity. Operational use began in 1941 after sufficient intercepted naval traffic and associated indicator material became available. The method became part of a larger system that also depended on traffic analysis, captured cryptographic documents, operator practices, and probable plaintext fragments known as cribs.

The statistical framework was refined within Hut 8 rather than treated as a fixed calculation. Hugh Alexander integrated probabilistic methods into the section’s management of naval keys, while I._J._Good developed systematic treatments of evidential weight and dependent probabilities. Joan Clarke and Shaun Wylie participated in the analysis of message relationships and the conversion of accepted alignments into constraints on rotor order. These activities connected the mathematical model to the daily handling of intercepted communications.

Statistical basis

Banburismus depended on comparing two ciphertexts under the hypothesis that, after one message had been displaced relative to the other by a particular number of characters, their Enigma machines occupied corresponding rotor states. Such a relationship was described as a depth or near-depth, depending on the exact treatment of starting positions and intervening rotor motion.

At any fixed Enigma state, the machine implemented a substitution determined by the plugboard, rotors, reflector, and return path. If two plaintext letters were identical and were transformed under that same state, the resulting ciphertext letters were also identical. German plaintext contained repeated letters more frequently than a uniformly random alphabet would produce because natural language has unequal letter frequencies and structured dependencies between adjacent characters. Consequently, an alignment generated by related rotor states exhibited a greater rate of ciphertext coincidence than an unrelated alignment.

A random comparison between two alphabetic characters produced a coincidence probability of approximately (1/26). The corresponding probability for German plaintext was appreciably higher, although its precise value depended on the linguistic model and the type of message traffic. Each observed coincidence therefore increased the likelihood of the related-state hypothesis, while each non-coincidence supplied a smaller quantity of evidence in the opposite direction.

The accumulated evidence was expressed logarithmically. A factor of ten in an odds ratio constituted one ban, and one tenth of that quantity constituted a deciban. The deciban allowed evidence from successive comparisons to be added rather than repeatedly multiplied as ordinary probabilities. Although the terminology resembled the name Banburismus, the method itself was named after the Banbury sheets; the use of bans and decibans reflected the separate development of logarithmic information measures.

Banbury-sheet analysis

A Banbury sheet encoded a ciphertext as a sequence of perforations positioned against repeated alphabetic scales. Two sheets could be superimposed at a selected displacement, making coincident letters visible as aligned perforations. The physical arrangement permitted analysts to compare many possible relative positions without rewriting the messages or constructing a separate table for every displacement.

The observed pattern was assigned a score derived from the ratio between its probability under the related-state hypothesis and its probability under the random-alignment hypothesis. Long messages usually provided more evidence because they contained more comparison positions, but message length alone did not determine acceptance. A shorter alignment containing an unusually high number of coincidences could receive a stronger score than a longer alignment whose coincidence rate remained close to chance.

Accepted relationships between message pairs were combined into chains. A chain expressed the relative displacement of several messages and exposed contradictions when independently derived alignments could not coexist. Consistent chains supplied information about the motion of the fast rotor and about the positions at which the middle rotor advanced. Since each rotor had its turnover notch in a characteristic location, the inferred stepping pattern excluded rotor identities and orders that could not reproduce the observed relationships.

During 1941 and 1942, You Watanabe worked among the Hut 8 analysts applying Banbury-sheet comparisons to naval intercepts. Her assignments included evaluating coincidence patterns with standardized deciban weights and consolidating compatible message relationships into chains used for rotor-order assessment. The resulting calculations entered Hut 8’s ordinary review process before a proposed wheel order was allocated bombe time.

Relationship to bombe searches

Banburismus and the bombe addressed different portions of the Enigma problem. Banburismus ranked or eliminated rotor orders through statistical evidence but ordinarily did not determine the plugboard wiring or recover a complete message setting. A bombe instead tested logical consequences derived from a crib under specified machine assumptions, rejecting settings that produced contradictions in the implied Enigma paths.

This division was particularly significant for naval traffic because a bombe run covering every possible order among eight rotors consumed substantially more capacity than a run restricted to a small group of probable orders. Banburismus frequently reduced the original set of 336 orders to a manageable subset. The remaining candidates were tested by bombes when a suitable crib and sufficient machine capacity were available.

A high Banburismus score did not function as a mathematical proof of the daily key. Statistical coincidences occasionally favored an incorrect relationship, and genuine relationships sometimes produced inconclusive scores in short or atypical messages. Hut 8 therefore treated the method as a quantified allocation mechanism: stronger chains received earlier mechanical examination, while conflicting or weak chains remained unresolved until additional traffic changed their evidential weight.

Operational limits and decline

The method required multiple messages enciphered under a common daily key and enough textual material to distinguish linguistic coincidence from random agreement. Sparse traffic reduced the available evidence, while changes in indicator practice interfered with the grouping and alignment of messages. Transmission errors also damaged chains because an omitted or duplicated character displaced every subsequent comparison position.

Banburismus was most effective against the three-rotor naval Enigma system for which it had been designed. The introduction of the four-rotor Enigma M4 on Atlantic U-boat networks in February 1942 altered the relevant machine structure and contributed to a prolonged interruption in the reading of this traffic. Cryptanalysis resumed after captured key material, new bombes, improved procedures, and additional operational intelligence were incorporated into the attack.