THE CODEBREAKERS

by DAVID KAHN

  The radio lull ended explosively at the start of 1916. This was the year in which the Germans oscillated wildly over the entire cryptographic spectrum in a frantic hunt for the ideal cipher. But the French kept up with them, and sometimes G.H.Q. received two or three solutions of a new problem within a few hours.

  Every possible weakness was exploited in these solutions. Particularly oful were stereotyped messages. “Night calm; nothing to report” appeared with what Givierge called “terrible regularity” in German transmissions. One command required regular morning reports from its units in line. When the cipher changed, the practice did not, and the French promptly pried open the cipher with the leverage of a known plaintext. The French turned to similar advantage the thoughtless German practice of checking out new systems by enciphering proverbs as test messages. The German version of “The early bird gets the worm” was a particular favorite—but it was the French who profited.

  Their familiarity with German habits of phraseology and transmission technique greatly helped. They had gained this insight during the very first days of the war, when the radio operators in General Georg von der Marwitz’ cavalry corps on the wheeling German flank became simultaneously intoxicated by their speedy conquests, overwhelmed by the volume of their traffic, and exasperated by the nuisance of ciphering. They began sending messages in clear. Soon, by a kind of cryptologic Gresham’s Law, everyone was doing it, while the French took copious notes. They bore down mercilessly on ciphering errors, studied captured notebooks with cryptographic worksheets, compared messages from different sectors that, individually, offered little but, conjointly, suggested much. They fished about wildly for keywords—and, given the German predilection for patriotic terms such as VATERLAND, KAISER, and DEUTSCHLAND, sometimes hooked the prize. They bombarded enemy trenches and feigned preparations for attacks just to get some badly needed probable words into enemy cryptograms. And, above all, they carved away at the ciphers with their keen, surgical minds, dissecting, discarding hypotheses, until at last they cut through to the heart of the system. Painvin in particular shone brilliantly in this pure cryptanalysis.

  The first of the new German systems appeared with the outburst of new wireless activity. The French high command believed that these signs presaged a new German attack, and Painvin and Olivari fell upon the intercepts. They quickly decided that half of them were fake—mere meaningless strings of letters. But what messages were the real cryptograms carrying? Within two weeks, they discovered that the system consisted of an interrupted-key Vigenère with key ABCD followed by a single columnar transposition. The key interruptions were controlled by the numbers of the transposition key. The system was an elaboration of the old ABC; they called it the ABCD. The plaintexts proved to be nothing but simple ciphering exercises, portions of communiqués, extracts from newspapers, even trigonometrical formulas. This showed that the entire radio busyness was a German deception, and the cryptanalysts thereby relieved the French staff of some of its worries.

  The cumbersome ABCD expired in April. It was replaced for the first time in German cryptography by pure substitution ciphers. These were numerous, but of two general kinds: monalphabetic substitution in which the choice of the 24 available alphabets was left to the encipherer, and polyalphabetic substitution with 12, 24, or 25 mixed alphabets. These grew ever more complicated, but as the development was progressive the French never lost cryptanalytic contact. Painvin solved one that spring thanks to a Bavarian prince’s telling his parents, the king and queen, that he had been wounded. The polyalphabetic systems culminated in one used between Berlin and Constantinople. It employed 25 alphabets, required 32 tableaux, and was so excessively complex that only cipher clerks comfortably ensconced in quiet, well-equipped headquarters offices could handle it. In fact, it was too elaborate, and the pendulum swung away from substitution to transposition. At the end of 1916, transposition messages again appeared in German military communications.

  By January, 1917, the French cryptanalysts recognized these as turning grilles. About all that these grilles have in common with the fixed concealment grille of Cardano is the name and the openings in the mask. The turning grille is usually a square sheet of cardboard divided into cells; one quarter of these are punched out in a pattern such that when the grille is rotated to its four positions, all the cells on the paper beneath will be exposed and none will be exposed more than once. A 6 × 6 grille might look like this:

  This is laid over a sheet of paper and the first nine letters are written through the apertures. Then it is turned 90 degrees, the next nine letters are written through the openings in their new position, and so on for two more turns. By then each of the 36 cells on the paper will have a letter inscribed in it, and the cryptographer can read it off in any pattern he chooses—usually by rows. Messages longer than 36 letters must repeat the process; in the last section of less than 36 letters, the unwanted cells can simply be blocked out.

  The Germans provided their signal troops with a variety of sizes for different length messages. Each grille had a codename: ANNA for 25 letters, BERTA for 36, CLARA, 49, DORA, 64, EMIL, 81, FRANZ, 100. These codenames were changed weekly.

  Grille systems are particularly susceptible to multiple anagramming—which is the general solution for transposition systems—because their sections are of necessity of equal length. But the system produces intriguing geometrical symmetries, and the French soon devised attacks exploiting this and other weaknesses. The grilles lasted four months.

  Britain, too, had her military cryptanalytic bureaus. But she had made no more preparations for them before the war than she had done for Room 40, and her Army cryptanalysts, expert though they became, never achieved the proficiency of the French.

  Her setup was essentially the same as France’s. The head organization, M.I. 1(b), was attached to the War Office. A field agency was established at British Expeditionary Force headquarters, and individual cryptanalysts were stationed with the several armies.

  M.I. 1(b) was still a small, four-man section—1(b)—of the Military Intelligence Division in December of 1915 when Malcolm Vivian Hay of Seaton was placed in charge. Hay, then 34, was the grandson of the second son of the seventh Marquess of Tweeddale and had succeeded to the Seaton Estates near Aberdeen when he was 2. After an education at Beaumont College and abroad, he returned to supervise his farms; he joined the Gordon Highlanders as a captain at the outbreak of war. He was machine-gunned at the Battle of Mons and was captured by the Germans when he was left on the field by the British retreat. Partly paralyzed as a result of his head wound, he was repatriated in February, 1915, as unfit for military duty. After learning to walk with the aid of a cane, he was promoted to major and given command of M.I. 1(b).

  He began at once to scour the universities for bright young men, preferably language scholars, to supplement the three original civilians on the staff: J. St. Vincent Pletts, a radio engineer from Marconi’s Wireless Telegraph Company; J. D. Crocker, a young Cambridge scholar, and Oliver Strachey of the Indian Civil Service, who liked cryptanalysis so much that he switched after the war from administering the East Indian Railway to codebreaking for the Foreign Office. Hay recruited a remarkable concentration of men who were later to achieve eminence, if listing in Who’s Who may be taken as an index. Among them were his chief assistant, John Fraser, 32, later professor of Celtic as a fellow of Jesus College, Oxford; Arthur Surridge Hunt, 45, then and later professor of papyrology at Oxford and one of the world’s most eminent authorities on ancient writing; David Samuel Margoliouth, 58, professor of Arabic at Oxford, later president of the Royal Asiatic Society and author of many works on Arabic literature and history; Zachary Nugent Brooke, 33, then lecturer in history at Cambridge, later professor of medieval history there and an editor of the Cambridge Medieval History; Edward Thurloe Leeds, 39, then assistant keeper of the department of antiquities of the Ashmolean Museum and, after the war, keeper of that first public museum in England; Ellis H. Minns, 42, then and later lecture
r in paleography at Cambridge, later knighted; Norman Brooke Jopson of Cambridge, 26, later professor of comparative philology there; George Bailey Sansom of the consular service, 33, later knighted and commercial counselor of the British embassy in Tokyo and author of a Historical Grammar of Japanese and of a standard history of Japan; and Henry E. G. Tyndale, 28, later housemaster of Winchester College, one of England’s great public schools, an avid mountaineer, and editor of the Alpine Journal and of the classic Whymper’s Scrambles Amongst the Alps. The chief himself, Hay, became well known as a historian, writing half a dozen major historical works (most presenting the Catholic viewpoint on controversial questions) and almost as many on other subjects. His first study, A Chain of Errors in Scottish History, concerning early church history, was violently denounced and extravagantly praised. But subsequent works, such as The Enigma of James II, were received with more moderate but more extended applause, and his later The Foot of Pride, an erudite examination of European anti-Semitism, was universally lauded.

  The staff of M.I. 1(b) was to number 84, including 30 women, by the end of the war. To shelter this growing organization, as well as to conceal it from the curious, the War Office requisitioned a largish private house at 5 Cork Street, several blocks from its own building in Whitehall and behind the fashionable Burlington Arcade. Hay immediately instituted a complicated entrance procedure that involved locking visitors in a room temporarily to prevent their wandering about the premises.

  Early in the war, the French had provided the English with keys and techniques for the German military ciphers, and with this help, M.I. 1(b) was soon passing valuable information to the army command. Eventually a pool of skilled cryptanalysts was built up, including one who was familiar with Turkish. Perhaps the most brilliant at Cork Street was Captain G. L. Brooke-Hunt of the Royal Engineers, who had served in the Indian Army.

  Among his most difficult problems was the “Für GOD” system, which was so-called because all messages in it bore that prefix to show that they were for the German wireless station whose call letters were GOD. These messages were sent irregularly about three times a week from POZ, the powerful German station at Nauen outside of Berlin. They began in 1916 and lasted until the fall of 1918, making the Für GOD the longest-lived German cipher. Because the dispatches bore no signature and no address beyond the call-sign, suspicion grew that the cryptograms concealed instructions to German secret agents.

  Brooke-Hunt solved the Für GOD early in 1917. It proved to be a poly-alphabetic system using 22 mixed alphabets and 30 incoherent keys of from 11 to 18 letters. The messages were numbered serially from January to December in each year and the keywords repeated in cycles of 30. The dispatches were transmitted by the political section of the German general staff to an expedition sent to North Africa under Captain von Todenwart to foment uprisings by the Arab population. Some of the messages were orders, but many forwarded reports of the slaughter of colonial troops on the Western Front as a result of alleged French placement of them in the most dangerous positions in the line. Von Todenwart was directed to spread these reports as anti-Allied propaganda.

  Among the messages were several arranging for a submarine to bring rifles and ammunition to Abd el Malek, a Moroccan nationalist. Hay was in the closest personal touch with Captain Hall at the Admiralty. Information was passed, wheels turned, air commands were notified, and shortly after the U-boat surfaced in the blue Mediterranean she had submerged again—this time involuntarily and for good, taking her cargo with her. Later in the war Brooke-Hunt read with mingled pleasure and regret a Für GOD message declaring that “For security reasons, U-boat arrival notifications will no longer be made.”

  Hay, who was admired by his subordinates as “a very good chief” (they later gave him a silver loving cup and a book of photographs with cryptographic inscriptions and affectionate remembrances), was given charge of constructing codes and ciphers for British forces early in 1917. He took his responsibilities seriously enough to make a visit to Cartier’s office despite his own disability, and later in the war his office sent representatives to the Near East to coordinate cryptologic security there.

  But M.I. 1(b) apparently had no hand in the development of perhaps the finest British cryptanalyst of the war. O. T. (for Oswald Thomas) Hitchings had been destined to be a schoolmaster like his father, but he loved music so much that he became an organist instead. Later he taught music in two preparatory schools, and while doing this learned French and German so well by correspondence that he won an honors degree in them from London University. In 1911 he went to Bridlington Grammar School as modern language master. Quiet, conscientious, he volunteered for the Army at the start of the war and went directly to France, where his knowledge of languages was put to use in the Field Censor’s Office. One day his colonel asked him if he would like to try solving the German messages that were being intercepted. He said he would, found he had a flair for the work, and by 1918, when he was 42, had risen to the rank of captain and the command of Intelligence E(c), 2d echelon—the Code and Cipher Solution Section of the British Expeditionary Force’s general headquarters.

  This section was located in Le Touquet, a Channel-side resort town not far from British G.H.Q. at Montreuil, probably for reasons of security. Here the serious, earnest Hitchings was assisted by a debonair, kilted Scot, Duncan Campbell Macgregor. Under them worked the cryptanalysts at the several army headquarters; at one of these an American visitor was astonished to see a German prisoner of war, still wearing his uniform, puzzling over the intercepts of his native land! Hitchings’ solutions were so extraordinarily valuable that one colonel exclaimed that he was worth four divisions to the British.

  With this superb background in cryptanalysis to instruct them, what systems did the Allies use? The British employed the Playfair with random keysquare. Its use extended even to Lawrence of Arabia. Behind the lines, the French corresponded in a four-digit superenciphered code; they changed it three times between August 1, 1914, and January 15, 1915. Series 65 of this code chiffré was a two-part code of about 2,300 four-digit groups. A tableau de concordance superenciphered number pairs into letter pairs with a straddling gimmick: the first digit was chopped off and enciphered separately, and the subsequent division into pairs straddled the gap between codegroups. This kept a codegroup from being always superenciphered the same way. A sample encoding and encipherment of the plaintext “The relief will take place tomorrow morning” in Series 65 would be:

  The remarkable French acuity in matters cryptological is nowhere better shown than in the instruction accompanying this code: “Exceptionally, if you do not have the time to encipher entirely, transmit in clear.” The French knew that partial encoding, which offered quick and easy entries into a code (“Colonel seriously 6386” could have but one meaning, for example), posed a danger to the compromise of all communications that a single cleartext message, which at best disclosed a single piece of information, did not.

  In the field, the French sometimes used a mixed-alphabet polyalphabetic with a running key. But the cipher they relied upon for three years was an interrupted columnar transposition that was, paradoxically, theoretically weaker than the German double transposition. It employed the usual transposition block with a key sequence in which the plaintext was inscribed horizontally. The vertical transcription, however, was preceded by a reading out of letters on certain diagonals. For example, with the message Enemy has brought up four howitzer batteries and three companies Stop We can hold but we need more fifty calibre machine gun ammunition Third Battalion (plus three nulls to complete the last five-letter group), and the key (the French used long ones) MADEMOISELLE FROM ARMENTIERES, with the rightward diagonals starting under 3, 5, 7, 8, and 10 to be taken off in that order, followed by the leftward diagonals under 16, 18, 21, and 26:

  The transcription begins with EAPCH and continues with BEHET. The leftward diagonals skip over any letter previously transcribed; thus, diagonal 21 would read TLB and not TDLEB. Similarly, the
vertical transcriptions ignore any letters taken by the diagonals: column 1 would read NRST and not NRSIT. The full transcription, which would naturally be divided into groups of five for transmission, is: EAPCH BEHET UOEA WNRN GDBHI YTII OETA TLB ZIOM NRST PRI BFI MOTO IAIR UAOA CNGA AM TU NM AEEA OPD RNBD OSR EESF TYN UHUL EEEN REUB HTWT TC HDAT FWNO IM SRCLI EE HMNC.

  The diagonals break up the columnar segments that the cryptanalyst juxtaposes and adjusts to solve uninterrupted columnar transpositions. But the diagonals constitute segments of their own, and the columns, though fragmented, keep their constituent letters together instead of scattering them, as does the double transposition. The cryptanalyst can seize upon these weaknesses to reconstruct the tableau. The task is admittedly more difficult than with an ordinary columnar transposition, but it can be effected with a single message far more easily than with the German system.

  Why, then, did the Germans not solve it for the three years that the French kept it in force?

  The reason is absurdly simple: Germany had no cryptanalysts on the Western Front for the first two years of the war.

  She had entered the war with no military cryptanalytic service. (An expected side effect appeared in the erratic development of German cryptography. The absence of the stabilizing influence of cryptanalysts resulted in the overcorrective swings from one field cipher to another in 1915 and 1916. The lack of cryptanalytic instruction also forced the Germans to attend the hard-knocks school of cryptography, learning through one painful experience after another the dangers of normal alphabets, patriotic keys, their inherent love of order, and the like.) But even if Germany had had well-trained cryptanalysts available at the start of the war, she would have had little opportunity to use them.