The ISK operation was an outstanding success. A number of amateurs in Bletchley, backed by a very few professionals, made a major difference to the Allied war effort - and helped to save D-Day.
RE
Intelligence from decrypts of the Abwehr (the German Secret Service) was largely responsible for enabling MI5 to control the entire German espionage network in the United Kingdom. It also enabled the organizers of the ‘Double Cross’ system of playing back German agents to be sure that they were successfully deceiving the Germans and, through strategic deception, played a major part in the success of the D-Day landings.
Radio traffic between various European capitals and Berlin, which had been intercepted – though spasmodically and in small amounts – from late 1939, was thought to be between outstations of the Abwehr and its headquarters, and had been diagnosed as having been enciphered on a four-wheel Enigma from its use of eight-letter indicators, such as GIWM XPEB, since that was a common practice when using the four-wheel commercial Enigma (see page 279, paragraph 3). The traffic was referred to Gordon Welchman in Hut 6 early in 1941, but although the number of signals intercepted had increased, following breaks into Abwehr hand ciphers in 1940, it remained much too insignificant to offer any prospect of evaluating the indicators, and as no cribs were available Hut 6 made no progress with the problem. Welchman, no doubt not being serious, asked me if sufficient information could quickly be got to break the machine if someone could get into an embassy for the purpose; I was discouraging, and mercifully no such tactic was tried.
As Dilly Knox had handed over his work on the Italian naval Enigma, he was asked to tackle the unsolved traffic around mid-1941 and, remarkably, had made great headway with the solution by October 1941. His note, dated 28 October, reporting his success to Alastair Denniston, the operational head of GC&CS, is endearingly eccentric and typically obscure; sadly, it ends with a scribbled postscript saying that he would welcome a discussion but would be away on 30 October – in fact his stomach cancer was to keep him away for the rest of his life. Ralph Erskine sent me a copy of the note in 1999 – until then I had no reliable information about how Dilly had solved the problem.
Dilly saw, as had Hut 6, that the number of signals on any one day was too small to make it feasible to evaluate the indicators (i.e. decipher the message settings from them), but he saw that if he could find two days where the same wheel order was used and such that the Grundstellung of one day could be got from the other by rotating each wheel and the reflector through the same number of places, he would in effect double the number of indicators available on one setting, because the letter pairings at each position of the Grundstellungen would be related by a QWERTZU … substitution (see page 281, paragraph 9). He therefore organized a search for two such days with the help of the Bletchley Park card sorting and tabulation section – this must have involved Dilly in some very persuasive negotiations, because the section’s resources were usually overloaded. The search was unsuccessful, which Dilly thought suspicious, though why is puzzling: it is doubtful that there were as many as a hundred days of traffic worth working on, implying fewer than sixteen on each of the six possible wheel orders. The probability that two days would have the desired relationship was 1 in 17,576 – but Dilly was disinclined to be troubled by probability calculations (an attitude inculcated in his lieutenants). Freeborn did, however, find a day on which the positions 1–5 and 2–6 fitted the bill – which Dilly reported as ‘finding what was wanted standing, Like the abomination of desolation, precisely where it should not – on a single setting’. He called the phenomenon a ‘crab’ on the basis that matters moved sideways. He also said that the discovery came as no surprise to him: presumably by this he meant that he was ready for something odd; if he had had the slightest suspicion that the machine had numerous wheel turnovers he would surely have had each day’s indicators carefully looked at, in which case crabs would have been found with little trouble, and Freeborn saved a tiresome task.
The crucial discovery of the crab was pure serendipity, but Dilly took full advantage of it: from the discovery he deduced that:
a) the Enigma had a QWERTZU… diagonal: the sequence from the top right to the bottom left of the rod square when the rod labels are in QWERTZU order (see Figure 17 and page 282, paragraph 8);
b) occasionally, between consecutive positions four places apart on the Grundstellung, all the wheels and reflector moved together;
c) hence the reflector moved during enciphering – a feature not previously encountered by GC&CS in an Enigma; d the wheels had numerous turnover positions, otherwise crabs would be very rare: this was also a feature not previously met with;
e) ‘as everything that has a middle also has a beginning and an end’ there would be positions at which all the wheels and reflector turned together without doing the same four places later – this he called a ‘lobster’, arguing that it was half a crab.
Dilly said that he condemned a crab as useless because, although it was a great help in finding the alphabets at the four positions of the Grundstellung, the alphabets could not lead to deductions about rods (see page 280, paragraph 8 (d)) because of the turnovers between each pair. Lobsters were useful because the QWERTZU … relation between the two alphabets at the lobster position greatly helped to discover them, so giving the alphabets four places away, which, with luck, would not be separated by a turnover.
At this stage Dilly organized a lobster hunt: he probably restricted the hunt to days with at least ten indicators intercepted; for each day ‘chains’ were made for places 1–5, 2–6 and so on; if a letter pairing was assumed in, say, place 1, the chains gave deductions about other pairings in 1 and 5, and if a lobster existed 1–2, there would be several pairings implied for 6; it was then easy to see whether these were consistent with the implications of the chain 2–6. Clearly the hunt involved a great deal of careful, tedious work, but Dilly announced with evident pride that ‘after two days Miss Lever,* by very good and careful work, succeeded in an evaluation which contained sufficient non-carry units to ascertain the green wheel’.
The evaluation of the first day showed that the operators (no doubt contrary to strict instructions) nearly always used Christian names, swear words, or obvious keyboard sequences as message settings, and as this greatly eased the evaluation for other days a fair number of solutions soon accumulated, from which followed recovery of the wiring of two other wheels, one (designated Blue) with 15 turnovers and the other with 17 (Red). With so many turnovers these recoveries were difficult, the Red especially so – it was only recovered by David Rees after a complex operation. Each evaluation produced seven consecutive places at which it was shown whether the wheel had a turnover or not: if a turnover the position was marked by +, if not by – so that the turnovers produced by an evaluation would be shown by a sequence such as – +++ – + –. With evaluations for several days these sequences could be fitted together in a process akin to dendrochronology (though much simpler!) to give the complete sequence of turnovers for the wheel, known as the ‘wheel-track’.
Having discovered the wheel wiring and wheel-tracks, it remained to discover how the letters on the wheel tyres related to the turnovers. This necessitated decipherment of a message, a formidable problem which Dilly tackled at home because of his illness. His reports do not say how he accomplished the break, and I was unable to ask him as I only followed David Rees on secondment from Hut 6 (permanent, as it turned out) at the beginning of November 1941. Dilly was helped by Margaret Rock, who stayed with him for some time at his home in Courns Wood, but I never asked her how the solution came about, possibly because she too was off sick for some months from December 1941. Undoubtedly Dilly used rods, since other possible attacks would have involved statistics, which he disliked. Analysis of the first letters in the intercepted messages established that most, if not all, began ‘NRX’ (‘Nr’ followed by ‘x’ as a separator), since N very rarely appeared in the first position, R in the second, or X in the third
. Dilly, or Margaret, would have noticed that the Green wheel rod square contained adjacent occurrences of RX on two rods (two letters on the same rod were called beetles by Dilly) and a N*X beetle. On a day with twenty or more messages – according to my recollection quite common in November, probably because more resources had been allocated to interception following Dilly’s break – it would be unlucky not to have messages starting at two of the beetle positions, and the message settings would confirm that their relative rod positions were correct; the beetles would be easily found by noting the indicators in which the bigrams associated with the beetles on the rod square appeared as the second two, or the first and third, letters. This process, repeated if necessary on two or three days with the Green wheel in the right-hand (RH) position, would fix the tyre in relation to the rod positions, which would be related to the wheel-track’s eleven turnovers by the evaluation of the indicators.
The Blue rod square had NRX on one rod, but no other relevant beetles; finding at the start of a message one of the trigrams in the same column of the rod square as the NRX beetle would make it highly probable that it gave the rod position at the start of the message. There were just three places on the Blue wheel-track which had no turnover in three consecutive positions of the wheel, so that study of the indicator evaluations for several Blue days would, by the process described, fix the tyre in relation to the wheel-track. The same procedure would, though with more trouble, fix the tyre in relation to the Red wheel-track, the Red wheel having a N*X and a RX beetle; Dilly, having sorted out the other two wheels, possibly used rods to read more text in one or two messages.
However Dilly went about the problem, Denniston informed ‘C, who was also the director of GC&CS, that ‘Knox has again justified his reputation as our most original investigator of Enigma problems … He read one message on December 8th. He attributes the success to two young girl members of his staff. Miss Rock and Miss Lever, and he gives them all the credit. He is of course the leader, but no doubt has selected and trained his staff to assist him in his somewhat unusual methods.’ Denniston implied that Dilly was being kind, but in fact he greatly valued the work of his two lieutenants, saying ‘Give me a Lever and a Rock and I will move the Universe.’ This reading of a message completed Dilly’s triumph and gave him the satisfaction of being brought into the mainstream again, having been distanced from the work of Hut 6 and Hut 8 (rightly so, because the crib/machine production line methods involved were not his style). Dilly’s note of 28 October 1941 said that he expected the recovery of keys to be laborious and he asked for the gradual return of the Cottage staff who had been seconded to Hut 8 during the slack period.
The correspondence that followed with Denniston harked back to the days of the first breaks of wartime Enigma when Dilly wanted to be involved with the intelligence side of codebreaking. ‘A scholar is bound to see his research through from the raw material to the final text,’ Dilly complained. Denniston’s reply on 11 November 1941 reveals their relationship with each other – also Denniston’s diplomatic forbearance!
If you do design a super Rolls-Royce that is no reason why you should yourself drive the thing up to the house of a possible buyer, more especially if you are not a very good driver … You are Knox, a scholar with a European reputation, who knows more about the inside of a machine than anyone else. The exigencies of war need that latter gift of yours though few people are aware of it.
Dilly was in fact terminally ill and never did take charge of the section which became known as ISK (Illicit (or Intelligence) Services Knox) to his great satisfaction; the name was highly appropriate as the Abwehr hand-cipher traffic had been broken in 1940 by Oliver Strachey, Dilly’s colleague from the early days of GC&CS, and his section was named ISOS (Intelligence (or Intelligence) Services Oliver Strachey). It was Peter Twinn, Dilly’s first colleague in the Cottage in 1939, who took over the running of the new ISK section early in 1942. Dilly remained at home in Courns Wood, with Margaret Rock liaising between him and Bletchley Park. He worked from his bed until he died on 27 February 1943, only getting up in order to receive the CMG ‘for services to his country’ from the Palace emissary.
ISK became operational early in 1942 and finally expanded into a staff of over one hundred in one of the new blocks. Four European Abwehr networks were attacked by ISK: two in the West and two in the East. As they all used different daily settings a considerable amount of work was involved. In spite of Dilly’s initial misgivings, evaluation of settings from doubly enciphered indicators proved fairly easy, and after straightforward rodding it was determined which was the RH wheel, the rod position at the start of the Grundstellung, and the wheel-track, which together fixed the RH wheel Ringstellung. The rod couplings for the RH wheel at the Grundstellung positions in effect gave parts of the alphabets produced by the other two wheels and reflector at those positions, and so enabled the middle wheel and its rod position to be determined, but probably not its wheel-track because only two or three turnover positions would be involved. This meant that determination of the Ringstellung depended on rodding out a few letters at the start of a message, which could prove troublesome.
Priority was given to the production of catalogues (on the lines developed by Gordon Welchman for the Railway Enigma), which provided a convenient and quick way of finding which settings of the reflector, LH and middle wheels gave prescribed pairings required by the rod position discovered for the RH wheel. When these catalogues became available, after three or four weeks, the day’s key was easily settled once the RH wheel and its rod position at the Grundstellung were found. Priority was also given to the conversion of Typex machines to deal with the work of deciphering messages after the day’s key was found, because the multiple turnovers of the Abwehr machine made deciphering by hand exceedingly slow and barely practicable without a vast staff. This was an inconvenient result from the multiple turnovers which, as just shown, were actually immensely helpful in working out daily keys because the wheel-track fixed the RH wheel Ringstellung in relation to the rods.
Very occasionally a day’s indicators proved obdurate. To deal with such cases Mavis Lever, in spite of my strong scepticism (long regretted), had charts made for NRXEINS etc., arguing intuitively that the correct rod position of the RH wheel would probably have a chart ‘click’. This scheme worked and proved invaluable when the double encipherment of message settings was stopped late in 1942, leaving the NRXEINS cribs as the only entry to the traffic. In 1943 ISK acquired a special four-wheel bombe (known as Fünf) to handle days when rodding failed. Charts and bombe together enabled nearly all Abwehr Enigma traffic to be deciphered until the end of 1944 when an entirely new machine, the Schlüsselgerät 41 (SG 41), was introduced.
Dilly had to start his attack on the unsolved Abwehr cipher machine from the initial assumption that a form of commercial Enigma was involved, since at that stage no other kind of four-wheel Enigma was known to GC&CS. The following notes, which relate mainly to the commercial machine, give some of the theory of the machine which he could use in testing that assumption.
FEATURES OF SIMPLE FOUR-WHEEL ENIGMA
1. The Enigma connected a typewriter keyboard to a bank of light-bulbs through a variable electric circuit, so that if a letter – A, say – was pressed a bulb – X, say – would light; the circuit was reciprocal, so that if X was pressed at the same setting A would light.
2. The variable circuit was provided by three wheels coaxially mounted with a reflector on the left and a fixed circular plate (the entry rotor – Eintrittwalze – or end-plate) on the right. Each wheel had twenty-six spring-loaded pins equally spaced around one face, each wired through the wheel to one of twenty-six small discs similarly placed around the other; the reflector had twenty-six pins placed like those on the wheels; and the end-plate had twenty-six discs, each wired to one keyboard letter and one bulb. Each wheel and the reflector carried a movable tyre or ring around its rim, lettered A–Z spaced to match the pins and discs; when mounted in the machine ea
ch could be set in one of twenty-six positions, each identified by the letter on the tyre which showed through a window in the machine’s cover. The tyre could be fixed relative to the body of the wheel or reflector by a clip fixed to the wheel’s body at one end and having a small pin at the other, which could be fitted into a hole in the edge of the tyre, identified by the adjacent tyre letter. The wheels could be mounted in the machine in any order, with the pins to the right, and when mounted were pressed together by a lever-operated spring, ensuring effective electrical contact between adjacent sets of pins and discs. When a keyboard letter was pressed, the right-hand (RH – i.e. that next to the end plate) wheel rotated anti-clockwise (viewed from the front right) one place; in each complete rotation of the RH wheel through twenty-six places the middle wheel would move one place, at a position determined by a notch on the RH tyre (called the turnover position); and the left-hand (LH) wheel moved one place for each complete rotation of the middle wheel. The reflector remained stationary during encipherment, but could be set manually.
3. To decipher a message the recipient had to know the order in which the wheels were mounted, the clip positions (Ringstellungen (ring settings) to the Germans), and the position of each wheel at the start of encipherment (the message setting). The message setting could not safely be sent in clear: one way of hiding it, often used by the Germans, was to encipher it twice at a specified basic setting (the Grundstellung) and transmit the eight resulting letters as the indicator. Cipher instructions for an Abwehr network gave a key for a period, usually a day, specifying the wheel order, Ringstellung and Grundstellung to be used. Each operator was usually allowed to decide the setting he used for each message; inevitably, instead of using meaningless sets of four letters as intended, many operators chose names, keyboard sequences, or swear words.
The Bletchley Park Codebreakers Page 34
