by John Keegan
Jackson was exceptionally arrogant. A Royal Naval Volunteer Reserve lieutenant, W. F. Clarke, belonging to the OB 40 staff, recorded that he “displayed supreme contempt for [our] work. He never came into the room during [my] time there except on two or three occasions, on one of which he came to complain that one of the locked boxes in which the information was sent him had cut his hand, and on another to say, at a time when the Germans had introduced a new codebook, ‘Thank God I shan’t have any more of that damned stuff.’ “3 There were many like Jackson, however, if not so bad, and it would take nearly a generation to pass before operations officers would begin to accept that most “raw” intelligence was only as good as the interpretation put on it, often best supplied by the intelligence officers who gathered it on a day-to-day basis.
OB 40 also had its admirers, and rightly so. It began to supply crucial information almost from the start, including forewarning of the raid on the English east coast towns of Scarborough, Hartlepool and Whitby on 16 December 1914.4 Had it not been for a visual signalling error committed by Beatty’s flag lieutenant—who was to repeat his mistake on three later occasions, at Jutland with disastrous effect—the Scarborough raid might have resulted in the destruction of the German battlecruiser force.5 OB 40 had only then been in existence since 8 November and had been brought into being because of an intelligence windfall. In late October the Russians had delivered to the British a copy of the main German naval codebook (SKM) and a collection of square-ruled charts, used to denote sea areas. They had been recovered from Magdeburg, a German light cruiser lost in the Baltic on 26 August. OB 40 subsequently acquired the codebook used for communication between German merchant and naval ships (HVB), found in a German steamer interned in Australia early in the war. Finally, it got possession of a codebook used by German senior officers (VB), allegedly dredged up in the nets of a British trawler off Holland on 30 November, at a spot where four German torpedo boats had been sunk on 17 October.6
It was with this material, and intercepts collected via a hastily established chain of coastal listening stations, that OB 40 went to work. They were aided by the Germans’ very free use of wireless—forced upon them in part by the dragging up of their oceanic cables by the British cable ship Telconia on 4 August 1914—but above all by the nature of the means the Germans used to disguise their signals.
Secret writing takes two forms, known to cryptologists respectively as codes and ciphers. Cipher is a method hiding meaning by altering the form language takes, either by “transposition” or “substitution.” Transposition, a technique so ancient that there is no record of its origins, works by changing the order of letters; the simplest system, familiar to any cipher-minded schoolboy, is to shift once along in the alphabet, so that A becomes B, B becomes C and so forth. “The cat sat on the mat” is thereby recorded as “UIF DBU TBU PO UIF NBU”; the result is unlikely to baffle an interceptor for any length of time. There are many ways of complicating a message in transposition cipher; one of the simplest is to run the letter group together—UIFDBUTBUPOUIFNBU—to disguise the word length, but it provides little protection. Another, more sophisticated, is to shift two or three or ten letters along in the alphabet; while straight transposition underlies the ciphering, however, the iron laws of “frequency analysis” will yield the solver a way in. The law of frequency reveals that, in English, E is the most commonly written letter, followed by A and so on. Frequency tables, known to all cryptologists, provide a ready means of decipherment. Frequencies are different in other languages—Z, rare in English, is common in Polish—but the tables cannot be defeated.
Not, that is, unless complexities are introduced. Cryptographers—those who write ciphers or codes—have devised many complexities. Perhaps the best known, and most difficult, is the alphabetical grid, which arranges the twenty-six letters of the Roman alphabet (reduced to twenty-five by combining the letters I and J) in a square five letters wide and five deep, and numbers the columns. If A is the first letter in the top left-hand corner, it is rendered as figures 11, and so on to Z as 55. At its most elaborate, known as a Vigenère, after its sixteenth-century French inventor, the square is twenty-six by twenty-six, presenting a frequency problem of great difficulty. It is not insurmountable, though it was long thought to be so.7
Further complications may be devised, particularly when cryptographers begin to use figures rather than letters in transposition. There developed, during the seventeenth century, a strange halfway house between transposition and its cipher alternative, substitution, in which, for example, King Louis XIV’s principal cryptographers, the Rossignols father and son, rendered whole words into mathematical figures. The technique had been anticipated by numbering common French “digraphs,” e.g., QU, OU, DE, but the Rossignol system, known as the Great Cipher, defeated everyone; it was only unlocked, long after the meaning of the messages in which it was written had ceased to have importance, at the end of the nineteenth century.
By then, however, cryptologists were on the brink of instituting a new cipher system altogether, employing full-scale mathematical “substitution” for letters. Mathematical substitution appeared to promise true impenetrability since by addition or subtraction, a numerical message could be so varied that a cryptanalyst—who attacks secret writing—would simply be defeated by time; but as long as the intended recipient possessed the “key” to understanding the chosen mathematics, it could be read at the other end.
Keys were the problem: how to ensure that senders and recipients possessed the same set, how to deny keys to the enemy? The simplest solution was to write the keys in a book, logically arranged, which could be owned by all legitimate parties. Codebooks were widely in use during the eighteenth century, if only to disguise the more important words in a message, for example, the proper names of people, places, ships and so forth, the rest being left in plain language. Major Benjamin Tallmadge, George Washington’s chief of intelligence after 1778, devised a codebook out of Entick’s Spelling Dictionary by taking from it the most frequently used words, numbering them in alphabetical or numerical order and adding random words for those not listed. He also chose sixteen numbers for key individuals and thirty-six others for cities or places. Tallmadge kept the original, sent another copy elsewhere and the third to George Washington. It cannot have disguised much from the British if a letter of 15 August 1779 is a fair sample: “Dqpeu [Jonas] beyocpu [Hawking] agreeable to 28 [an appointment] met 723 [Culper Jun.] not far from 727 [New York] and received a 356 [letter].”8
This amateurish example discloses the principal weakness of a codebook: that, by collecting the words used, from messages intercepted, parts of the book can be reconstructed; if enough material passes through the hands of the enemy, it can be reconstructed in its entirety.
An apparent safeguard is to avoid the use of the alphabet altogether and employ only figures, singly or in groups. By the beginning of the nineteenth century the British were doing just that. A message from William Drummond, British emissary to Denmark, to the Foreign Secretary, Lord Grenville, in 1801, on the eve of the Battle of Copenhagen reads, in its last sentence, “3749 2253 529 2360 1268 2201 3356,” which stands for “Count Bernstorff does not even affect to conceal his alarm and inquietude.”9 The protection, however, is not as great as seems, even though, in the original, there is no indication of sentence length, nor do the groups betray word length. But by painstaking accumulation, again, watching for repetitions, and by guesswork, the codebook can be reconstructed and meanings deduced.
From the use of all-figure codes, it was but a short step to a much more secure system, technically known as “super-encipherment.” It employed two—or more—keys: the codebook itself and a system of figure alteration, by addition or subtraction. Since the groups thus altered did not coincide with those in the book, and did not obviously repeat themselves, retrieval of meaning became much more difficult. Yet not impossible: there was an underlying logic, supplied by the second key, which might be established by mathema
tical analysis. German diplomatic telegrams were, during the First World War, commonly super-enciphered. Oddly the most famous, the Zimmermann telegram, was not. Broken by Room 40, its contents—which encouraged Mexico to attack the United States—prompted President Wilson’s declaration of war on Germany.
Many other methods of complexifying secret writing had been devised by the beginning of the twentieth century, many of them variations on the Vigenère square. The most ingenious was invented by an American army officer, Major Joseph Mauborgne, in 1918. It came to be known as the “one-time pad” and was indeed unbreakable. A Vigenère square was constructed in two copies, one held by the sender, one the receiver. It gave the key to the enciphered message; once used by both parties, it was destroyed. One-time pads protected messages absolutely because the coincidence between cipher and plain text was entirely random and the absence of repetition, assured by destruction, forbade all chance of frequency analysis, or any other method favoured by cryptanalysts.
The one-time pad suffered, however, from a disabling defect. To be useful, pads had to be distributed on a very large scale and had to be identified, so that sender and recipient knew that they were working on the same document. The generation of random numbers on a large scale is not a simple matter either; deliberate attempts to randomise will inadvertently follow patterns, the more conspicuously the greater the pace and volume of output, while large-scale distribution in real time poses insurmountable logistic difficulties. Any persuasive solution to the problem of randomising and distribution was, therefore, likely to be enthusiastically welcomed in military circles everywhere.
THE ENIGMA MACHINE
It was to be supplied by a German inventor, Arthur Scherbius, who in 1918 set up a small engineering business to produce and market inventions. One of the ideas he took up was for a machine that would encipher—but also decipher—automatically. Cipher machines were not a new idea; simple versions had long existed. One, indeed, had been invented by Thomas Jefferson, polymath and third president of the United States. It consisted of thirty-six discs, separately rotatable about an axle, on the rims of each one of which were engraved the letters of the alphabet in random sequence. The sender enciphered by turning the discs to produce a plain text (which could not, of course, be of more than thirty-six letters, though it might be of fewer). He then sent another row of letters to the intended recipient. The recipient turned his discs to replicate the message, which was a garble, and then examined all the other rows of letters. One was the plain text. Security was provided by arranging the discs in a different sequence on the axle, the change of order being preordained and known only to sender and recipient. Had the order of discs not been variable, messages would have succumbed quite quickly to frequency analysis; as it was variable, giving thirty-six possible orders, a number having forty-eight digits (36 3 35 3 34 3 33 . . .), messages were effectively irretrievable in the pre-computer age.10 In the period 1910–20, several attempts were made to mechanise the rotating disc principle though none achieved commercial success. Nor, at first, did the Scherbius disc machine, when offered for sale under the trade name Enigma in 1923. In the later twenties, however, Scherbius managed to interest the German armed forces in Enigma. Models were bought and adapted and in 1928 the German army began to use it for all secret communication susceptible to interception, which effectively meant radio messages. It was also taken into service by the German navy.
What particularly attracted the German forces to Enigma was a feature unique to the Scherbius system, the “reflector” disc, which equipped the machine to work both to encrypt and decrypt; a message encrypted on one Enigma machine would, when entered in encrypted form on another Enigma machine set up in the same way, yield the plain text automatically. That feature eliminated the need to decrypt by separate process, a tedious and time-consuming business. In that sense, the Enigma was an early example of the “on line” machine (though it was emphatically not a computer, but an electromechanical switching system).
Enigma had other characteristics making it attractive to military signal services: compactness and portability. Outwardly it resembled a portable typewriter of the period, with a typewriter keyboard, in the military version originally arranged alphabetically rather than in the QWERTY order, and a strong carrying case; there were no numerals, all numbers having to be spelled out. It was normally powered by dry-cell batteries.
The chief virtue of Enigma, however, lay in its ability to multiply possible encryptions by an order of magnitude so large as to defy decryption by an outsider in any practical dimension of time; estimates of how long it would take mathematicians to break an Enigma encryption by brute calculation vary, but the Germans themselves believed that the lifetimes of thousands, perhaps millions of mathematicians, working without sleep, would not suffice to decrypt a single message. Enigma was supposed to have complicated the making of the “key,” which is the heart of the cipher system of secret writing, to a degree lying beyond the power of human intelligence to produce a solution.
The purpose of the key is to disguise letter frequency and to multiply, to as near infinity as possible, the number of mathematical attempts necessary to establish a frequency table. The Vigenère square was one method of lengthening a key; there are many others, including the use of a common text, such as a word possessed by both users. The principle, however, remains constant: to make the key so long, consistent with convenient decipherment, as to defy mathematical process. Total mystification, unless by the one-time pad, is never to be achieved; the key has a logic and is therefore retrievable by reason; the object is so to overload the powers of reason as to defeat it in real time, indeed in any sort of human time at all.
Enigma appeared to do exactly that. Its electromechanical switching process was entirely logical; but unless the steps by which it worked were understood, and unless the basis on which the switching was started was known, then indeed the mathematics of its decryption became insurmountable.
The steps and the starting were separate from each other: the first was intrinsic to the machine, though variable within definable limits, the second was illimitable, in theory at least, being chosen by human decision.
The intrinsic characteristics of Enigma could produce five variables, most dependent on its discs: (1) the internal wiring of the discs, (2) the choice of discs, (3) the arrangement in order, from right to left, of the chosen discs, (4) the alteration of the disc rims, (5) the “plugging” (steckerung) of the discs from one to another.
Each Enigma disc, which was removable, had two faces, with fifty-two contact points for the letters of the alphabet; the right-hand face had twenty-six transmitter points, the left-hand face twenty-six receptor points; the interior of the disc wired the transmitter and receptor points together in a secret way.
When the key on the typewriter keyboard was depressed it sent an electric pulse through the right-hand (fixed) disc to the right-hand face of the first rotor. By internal wiring, that rotor transformed the impulse from, say, A to B on the disc’s left-hand face (in fact, the course of the wiring; something much more complex was expected by Germany’s enemies). The right-hand face of the second disc picked up the pulse and transmitted it by internal wiring to its left-hand face; the third rotor then worked similarly. When the pulse left the third rotor it was picked up by the fourth (fixed) reflector disc and sent back again along the same route as it had been received. With this difference: because each rotor was notched to turn over when it had received twenty-six pulses, the returning signal would find the right-hand rotor in a different position, by one letter, on its first journey, the second letter in a different position (by one letter) on its twenty-sixth journey and the third letter again in a different position, by one letter, on its six hundred and sixty-sixth (26 3 26) journey.
The eventual destination of the pulse was an electric bulb, each representing a different letter of the alphabet; as each lit up, in sequence, on the receiving machine, the bulbs would reveal the plain-text message. Befo
re the pulse reached a bulb, however, it went through another multiplying process; at the end of the return journey, it moved to a “plug” board, resembling that of a manual telephone exchange, on which six letters were plugged to another six (the number of plugs was later increased): A to E, for example, and G to T and so on; the pluggings were altered according to instructions for monthly, weekly, daily and eventually twice-daily use.
Thus the intrinsic complexity of Enigma. It was enlarged by human alterations. In the original version there were only three rotors.11 Part of the procedure laid down for Enigma’s use, in frequently changed instructions, was to alter the order in which the rotors were arranged in their slots. Finally, each of the rotors had on its outer rim a rotatable ring, often described as the “tyre on the wheel,” which would be moved to any one of twenty-six alphabetical positions. When setting up the machine for use, the operator moved the rim to a position laid down in instructions. The number of variables with which a cryptanalyst was confronted was therefore as follows:
