This now became my chief task, and Mr Backhouse left me more and more to myself, though he constantly checked and verified my work. What I found difficult to believe was that a book so regularly constructed would have ten decode sections on each page, since the various groups for each word appeared to be randomly selected within the hundred possibilities. I suggested therefore that the book was in fact so constructed that it would be easy to find where on the page each numerical group might be located, so that there would be no separate section for decoding. But Mr Backhouse would not hear of this.
My first success was to work out that each ‘page’ had the numbers divided into two classes. Those of the first class always preceded those of the second, so that each ‘page’ represented a double-page spread. All the groups on the left-hand side and all on the right shared a common feature. On the left-hand pages the first and last digits of every group were both odd or both even; on the right-hand pages, if the first digit was odd, the last was even, and if the first was even, the last was odd. The rule could be expressed by saying that if the sum of the first and last digit of a group was even, it belonged on the left-hand page; if odd, on the right-hand page.
This suggested that the arrangement of the numbers in each column, now of fifty, not a hundred groups, was also methodical. We had established the value of enough groups in the 900, 800 and 700 columns to be able to see the numerical connections between the groups with the same meanings in different columns. By this means I discovered that there were only three arbitrary factors to be worked out for each column: the starting point, whether the next group was that following in ascending or descending order, and a break point somewhere in the columns. Thus if the 900 column began with 939, the group for the next line would be either 941 or 937. The sequence then continued until it reached 999 or 901 or the breakpoint occurred. If the break came at 963, the next group would be either 999 and continue descending to 997, 995, etc.; or it would be 901 followed by 903, 905, etc.:
939 995
941 993
943 991 and so on to
945 983
947 and so on to 981
961 965
963 967
999 969 and so on to
997 979
Having established this principle, I was able to reconstruct a whole column by identifying only three or four groups in it. It was hard work convincing Mr Backhouse of this, but eventually he accepted some decodes on the basis of this method. Unfortunately the traffic dried up before we had been able to glean intelligence of any value. Since it was used in the Italian-held Dodecanese islands, it would have been invaluable if Allied operations had been launched in that area. As it happened, the decision was made to invade Sicily and then Italy, leaving the Italian overseas garrisons to their fate.
However, long before that date, in September 1942, a signal from Admiralty in London announced that Able Seaman J. Chadwick was promoted Temporary Sub-Lieutenant (Special Branch) RNVSR. Before Commander Murray left, he had told me that he had recommended me for promotion, but I hardly expected this. From what I could piece together afterwards I suspect the story went something like this. Having initiated me into the work and discovered my aptitude for it, my superiors suddenly realized that the material I was handling was officially classed ‘Officers Only’. On enquiring whether I could be made an officer, the Commander was told yes; I should be sent back to England for training and would be returned to service in Alexandria in about six months. This provoked the famous Murray temper, and as a result of the explosion the C-in-C was compelled to ask Admiralty for an immediate commission so that I could remain on the spot. The change of status brought about a very welcome improvement in my living conditions, but made not the slightest difference to the work I was doing.
About the same time we were joined by another new recruit, F. W. Ponting, a Sub-Lieutenant of the Fleet Air Arm, who had been trained as aircrew but to his immense disgust was then grounded for medical reasons. He had been an undergraduate at Christ Church, Oxford, reading mathematics, so he proved adept at the procedures needed to break recipherment keys. After the war he became a lecturer in mathematics at the University of Aberdeen.
Our best success was not the result of our own work, but of the capture of an Italian submarine named Perla with all its code-books. For about six weeks we were able to decipher and read every signal sent to its submarines by the Italian Admiralty in Rome. Unlike their German allies, but like our own, the submarines were instructed to keep radio silence while on patrol, so there were no signals from them. The recipherment key changed too frequently to be broken even with the benefit of the code-book, but at least we were able to prove that traffic analysis as applied to this code was useless, since on any one day the proportion of dummy messages varied from zero to 100 per cent. The information on patrol areas gleaned from these signals was of course very useful in keeping our shipping out of trouble.
The result of the decision to by-pass the area of our interest was that traffic declined, and I found myself under-employed. Mr Backhouse had no work to give me, and since I declined to remain idle for about half my time, I cast envious eyes on a pile of intercepts which was growing untouched in my office. They were received regularly from Malta, where there was a unit intercepting Italian naval radio telephone (R/T) signals. With some difficulty I prevailed on Mr Backhouse to allow me to take a look at this traffic. His reluctance was due to the fact that his orders from Bletchley were to leave this to them, but I was able to point out that there was a good naval tradition of ignoring orders when they seemed inappropriate to the man on the spot. I had no illusions that this traffic would contain any intelligence of value, but I thought that any fresh examples of naval signals might offer parallels which would make it easier to understand our assignment. So in my spare time I started to analyse these R/T signals.
Each began with a prefixed word: DRAPI, GIOVE or DELFO. It was soon clear that GIOVE and DELFO were alternative indicators for one code, but that DRAPI indicated a different one. Since there was less traffic in this, I put it aside and concentrated on the more frequently used code. I had hardly begun the indexing when news reached us that a sweep by units of the Western Mediterranean Fleet from Gibraltar had captured a small Italian auxiliary vessel, together with its documents. These revealed the existence of a chain of observation posts on fishing boats in positions at sea to cover the approaches to the Italian mainland and give warning of air raids. Among the documents was the code for the western chain with the prefix DRAPI. But instead of immediately circulating these documents to other intelligence units, the officer in charge at Gibraltar decided to have them translated, and he then sent us a copy of his translation. It is obviously ludicrous to try to translate a codebook, since without context the meaning of many words would be ambiguous. So the first thing I had to do on receiving this was to translate it back into Italian, guessing as best I could what the original was which had given rise to the translated word. I could then decode the traffic we had.
As I expected, there was nothing in it of interest. But the routine signals provided an easy key for the solution of the other code (GIOVE/DELFO) for the eastern chain, covering the straits of Otranto and the extreme south of the Italian peninsula. Therefore I was quickly able to read the routine traffic, but my attention was caught by a very long and obviously exceptional message of quite recent date. It used a number of groups which did not appear elsewhere in the traffic, so only a partial decode was possible. However, after working on it with Mr Backhouse we were able to translate enough to establish two facts. It dealt with a sunken ‘enemy’ (i.e. Allied) submarine. Now a British submarine was overdue from a patrol in this area, and had been presumed lost. Here was the first confirmation that it had been sunk, not far from the big Italian base at Taranto. Secondly, it seemed that the reason for the signal was to inform all units at sea that efforts were being made to salvage the vessel. This was of major importance, since if successful the enemy might well recover a copy of
the British submarine code. Mr Backhouse reported this immediately, and a signal was sent to Admiralty in London relaying these facts.
When it was received, an enquiry was made of Bletchley Park why the news had come from Alexandria and not the home station. As I learned much later, there were some red faces in the naval section. After the war, when I had resumed my studies at Cambridge, I was sent for a supervision to Patrick Wilkinson at King’s; to my astonishment he greeted me with the words GIOVE, DELFO. It took me some time to grasp what he was talking about.
The Italian surrender in September 1943 brought our work to an abrupt end. There were a few days immediately afterwards when a transmitter in Rhodes started sending en clair, and we were able to submit translations. What happened was that Rhodes then had a garrison of about 25,000 Italians and 5,000 Germans. The Germans had an air base on the east coast, from where their bombers could reach Egypt. But they succeeded in disarming and rounding up the much larger Italian garrison, no doubt by this time thoroughly demoralized. The naval personnel where the transmitter was located were isolated and sent out appeals for help; the Germans apparently adopted salami tactics and came at intervals and took away a few men at a time. The last message declared that there were only five of them left, and when the transmitter went off the air that would be an indication that they too had been taken. We were unable to do anything to help them.
However, elsewhere in the Dodecanese there were no Germans, and the Italian authorities at the other naval base in Leros, much further north, offered to co-operate with the Allies. Contact was established and a tiny British force was sent to stiffen the Italian garrison. But the only means by which the Navy could transport them there was by destroyers, each carrying only a small quantity of men and equipment. The destroyers sailed from Alexandria about 1600 hours and proceeding at top speed reached the straits leading to the Aegean after dark, arriving in Leros around 0200. They had two hours to unload, then had to sail again and take refuge for the following day in a bay on the Turkish coast. Here they were protected by the conventions governing territorial waters; but the German Air Force of course located them, and then amused themselves by making dummy attacks, hoping the ships would open fire, and thus expose themselves to real attack. Then the following night they had to make good their escape back to Alexandria. Several runs were successfully completed, but on one occasion the Italians were only informed of the arrival of the British destroyers after they had sent out their squadron of motor torpedo boats to patrol the approaches. They therefore sent them a signal giving the time of arrival and route to be followed by the British ships. They had taken the precaution of making a small change in the code (the special MAS code), but it took me less than half an hour to penetrate this, and I have no doubt that the German naval codebreakers were equally efficient.
As a result of this incident, I volunteered to go on the next mission to act as liaison with the Italian Navy in Leros, in the hope of preventing any further breaches of security. My suggestion was rejected, and I was told brutally that my superiors did not mind if I were killed, but they were unwilling to take the risk of my being taken prisoner. Although I was never on the Ultra list (people receiving high-grade signals intelligence), I had inevitably gathered a certain amount of information about it, and with hindsight this was a wise decision. But at the time I was annoyed at not being allowed to resume a more active form of service.
In any case, the next mission was the last. A force of four destroyers ran into a newly laid minefield off Kalymnos; one sank and another had to be beached on the Turkish coast. At the same time, against all expectations, the Germans in Greece managed to scrape together an invasion force and transport them on caiques and other small craft to Leros, so the operation was short-lived. It was as well I was not allowed to go.
After the Italian surrender I was for a time employed on general intelligence duties in the naval headquarters, but at the beginning of 1944 I was sent home by air from Cairo to London and, after my first leave for three years, started on a Japanese language course at Bletchley Park. This was run by a diplomat named John Lloyd, who had studied the language in Japan before the war; it was specifically designed to teach us to read written Japanese naval documents. Our main textbook was a photostat copy of the captured logbook of a destroyer squadron. Being handwritten it was a severe test of our ability to recognize characters. I was billeted in Bedford and had to travel to Bletchley by train each day; the work was arduous and the hours long. At the end of six months we were given an examination, and I was one of the two best students on the course; the other was another classics scholar. As a result I was assigned, not as I hoped and expected to cryptography, but to the very specialized unit which had the task of translating the decodes of the Japanese naval attaché machine cipher, JNA-20. We had no contact with the experts who had broken the system, a variant of the Japanese Purple machine. We merely received the signals as processed by a party of Wrens, who had machines for the purpose.
The personnel of this unit were most unusual. Nominally in charge, though he was frequently absent for unexplained reasons, was Lieutenant-Commander P. L. Nichol, RN, presumably retired, but rumoured to be employed by some branch of the Secret Service. How he had acquired his knowledge of Japanese I never discovered. The other two, named Watts and Pickles, were both diplomats who had served in Japan, and so had a good command of the language. So I found myself, the veriest beginner, in the company of three of the very scarce experts in the language. Naturally I was given all the uninteresting messages to translate, thus allowing the experts to work on the important ones.
At this time there were only two sources in Europe from which messages were sent to Tokyo: Berlin and Stockholm. It was my task to translate the dull bits of information gleaned by the man in Stockholm, very often by reading English newspapers and magazines. He was a regular subscriber to the periodical called Aeroplane and sent his masters any details about naval aviation which had been passed for publication by the British censor. When I asked why we should translate such material, I was told that it might be of interest to see what he had been told to look out for.
The real gems, of course, came from Berlin. The naval attaché was given many of the secret reports prepared by the German Navy on all kinds of technical matters. These were carefully translated into Japanese and a digest was sent to Tokyo in a long series of signals. We were well aware that some of the technical information was of the highest value, both in disclosing the research activity of the enemy, and telling our boffins how far the Germans had been able to discover the secrets of our latest weapons. I remember one particularly difficult set of messages, which I was allowed to handle, dealt with the details of the latest Allied airborne radar, which had been installed in an American bomber brought down sufficiently intact for the German experts to be able to study its construction. The technicalities were almost wholly incomprehensible to anyone without a good background in advanced radio engineering, and we eventually succeeded in obtaining a copy of the manual for this set, but even with this crib it was still very hard to make a good translation. Fortunately we were relieved from this need by the end of the war in August 1945.
I can recall two technical subjects – though I did not do the translation myself. They gave an account of the German experiments in connection with the design of acoustic mines; they had laid out a pattern of microphones on the sea-bed in relatively shallow water, and then run a variety of ships over them at different speeds, recording the volume of propeller-noise at different ranges. All this information was contained in our signals. We were told that if the Allies had performed such an experiment, it would have cost (in those days) well over £100,000 (in present values several millions). All this information was presented to our experts for next to nothing.
Even more vital was the account of the new generation of U-boats the German Navy was building in 1945. These were of revolutionary design, since they were the first submarines with a high underwater speed. The messages ga
ve full details of their performance. Fortunately very few of them had become fully operational by the time of the German surrender; but if we had had to deal with them at sea, it would have been very valuable to know their capability so exactly.
We occasionally had items of more general interest, when the officer in Berlin was able to pick up information which related to Germany’s ability to carry on the war. During the winter of 1944–5, the Japanese Ambassador called in his military and naval attachés for a serious appraisal of this subject, and after an exchange of information and views each sent his own report to his masters in Tokyo. The naval attaché’s version passed through our hands, and must have been of vital interest to the strategists directing the Allied war effort.
In order to encipher Japanese by machine it was necessary first to write the text phonetically in the roman alphabet, something no Japanese would normally do. There was an official system of romanization, called romaji, which differed from the more common one, as for instance in the Kenkyusha Japanese-English Dictionary which we regularly used. However, once you got used to it, this was no obstacle, any more than the special conventions adopted to help disguise the patterning of texts. The real problem lay in the technical terms, most of which were of too recent invention to appear in the dictionary. The Japanese were able to devise their own terminology by creating new compounds out of Chinese characters. Thus to describe a radar set, they came up with a new word dempatanshingi. This was written with five characters: DEN – ‘electric’, HA - ‘wave’, TAN - ‘search’, SHIN – ‘find’, Gl - ‘apparatus’. It was easy enough to work out the meaning of such a compound, provided you knew which the characters were. The trouble was that when written phonetically many characters had the same phonetic form, so each syllable might represent any of several characters, and in the worst cases there might be as many as twenty or thirty alternatives. For instance, DEN literally means ‘lightning’ but can be used to mean ‘electricity’, and could equally well be translated as ‘rice-field’.
The Bletchley Park Codebreakers Page 13
