The Bletchley Park Codebreakers (47 page)
Read The Bletchley Park Codebreakers Online
Authors: Michael Smith
The delays that cost the NPL the race with Manchester were not of Turing’s making. It had been agreed between the NPL and Dollis Hill in February 1946 that a team under Flowers’ direction would carry out the engineering work for the ACE, and in March 1946 Flowers said that a ‘minimal ACE’ would be ready by August or September of that year. Unfortunately, Dollis Hill was overwhelmed by a backlog of urgent work on the national telephone system. As Flowers said, his section was ‘too busy to do other people’s work’. In February 1947, Turing suggested that the NPL set up its own electronics section in order to build the ACE. This was done but, sadly, inter-departmental rivalry hindered the work, and in April 1948 Womersley reported that hardware development was ‘probably as far advanced 18 months ago’. Meanwhile, in the autumn of 1947, Turing retreated in disgust to Cambridge for a year’s sabbatical leave, during which he did pioneering work on Artificial Intelligence. Before his leave was over, he lost patience with the NPL altogether and Newman’s offer of a job lured a ‘very fed-up’ Turing to Manchester University. In May 1948 Turing was appointed Deputy Director of the Computing Machine Laboratory (there being no Director).
Newman had laid plans for his Computing Machine Laboratory following his appointment at Manchester in September 1945, applying to the Royal Society for a sizeable grant in order to develop an electronic stored-program computer. This was approved in July 1946. Newman introduced the electrical engineers Frederic Williams and Tom Kilburn – newly recruited to Manchester University from TRE – to the idea of the stored-program computer. Williams and Kilburn knew nothing about Colossus. At TRE they had worked
during the war on radar and they were expert with electronic pulse-techniques. At the time he left TRE, Williams was developing a method for storing pulse/no pulse patterns on the face of a cathode ray tube – an idea that, with Kilburn’s help, was rapidly to lead to the type of high-speed random access memory (RAM) known as the Williams tube.
Williams’ description of Newman’s Computing Machine Laboratory is vivid:
It was one room in a Victorian building whose architectural features are best described as ‘late lavatorial’. The walls were of brown glazed brick and the door was labelled ‘Magnetism Room’.
Here, Kilburn and Williams built the world’s first electronic stored-program digital computer, the ‘Manchester Baby’. As its name implies, the Baby was a very small machine. The first program, stored on the face of a Williams tube as a pattern of dots, was just seventeen instructions long. It was inserted manually, digit by digit, using a panel of switches.
Once Turing had arrived in Manchester, he designed the input mechanism and programming system for an expanded machine and wrote a programming manual for it. This expanded machine was known as the Manchester Mark I. At last Turing had his hands on a stored-program computer. He was soon using it to model biological growth, a field nowadays known as ‘Artificial Life’. While the rest of the world was just waking up to the idea that electronics was the new way to do binary arithmetic, Turing was talking very seriously about programming digital computers to think.
At the time of the Baby machine and the Mark I, Kilburn and Williams, the men who had translated the logico-mathematical idea of the stored-program computer into hardware, were given too little credit by the mathematicians at Manchester. They were regarded as excellent engineers, but perhaps not as ideas men. This was unfair, but now the tables have turned. During the official celebrations of the fiftieth anniversary of the Baby, held at Manchester in June 1998, Newman’s name was not so much as mentioned. Fortunately the words of the late Williams still exist on tape:
Now let’s be clear before we go any further that neither Tom Kilburn nor
I knew the first thing about computers when we arrived in Manchester University … Newman explained the whole business of how a computer works to us.
Newman had played a crucial role indeed in the triumph at Manchester. Through him, both Colossus and Turing’s abstract universal computing machine of 1936 were vital influences on the Manchester computer.
Meanwhile, momentum was gathering on the other side of the Atlantic. After visiting America in January 1947, Turing reported that the ‘number of different computing projects is now so great that it is no longer possible to have a complete list’. The most visible players were Eckert, Mauchly, and von Neumann. Eckert and Mauchly had entered into a contract with the US Army Ordnance Department in June 1943 to build an electronic machine, the ENIAC. Construction got under way at the Moore School of Electrical Engineering at the University of Pennsylvania. Von Neumann was involved in the Manhattan Project at Los Alamos, where human computers armed with desk calculating machines were struggling to carry out the massive calculations required by the physicists. Hearing about ENIAC by chance, he saw to it that he was appointed as a consultant to the Eckert–Mauchly project. By the time he arrived at the Moore School, the design of the program-controlled ENIAC had been frozen in order to complete construction as soon as possible. Programming consisted of re-routing cables and setting switches. Von Neumann brought his knowledge of Turing’s 1936 paper to the practical arena of the Moore School. Thanks to Turing’s abstract logical work, von Neumann knew that by making use of coded instructions stored in memory, a single machine of fixed structure could in principle carry out any task that can be done by a human computer. It was von Neumann who placed Turing’s abstract universal computing machine into the hands of American engineers.
After extensive discussions with Eckert and Mauchly, von Neumann wrote the ‘First Draft of a Report on the EDVAC’, describing a stored-program computer. The first draft (not quite finished) was circulated, bearing only von Neumann’s name. Eckert and Mauchly were outraged, knowing that von Neumann would be given credit for everything in the report – their ideas as well as his own. There was a storm of controversy. As a result, von Neumann abandoned the
proposed EDVAC and in 1946 established his own project to build a stored-program computer in the Institute for Advanced Study at Princeton. Von Neumann gave his engineers Turing’s ‘On Computable Numbers’ to read. The completed machine, with a high-speed memory consisting of forty Williams tubes, was working by the summer of 1951. It had approximately the same number of valves as Colossus II. Known as the IAS computer, this was not the first of the various stored-program computers under construction in the US to work, but it was the most influential, and served as the model for a series of what were called ‘Princeton Class’ computers.
Von Neumann is sometimes falsely described as the ‘inventor of the computer’ and the ‘inventor of the stored-program concept’. Books and articles that purport to tell the story of the stored-program computer sometimes place von Neumann centre stage and make no mention of Turing. Von Neumann himself, however, repeatedly emphasized the fundamental importance of Turing’s ‘On Computable Numbers’. For example, in a letter to the mathematician Norbert Wiener, von Neumann spoke of ‘the great positive contribution of Turing’, Turing’s mathematical demonstration that ‘one, definite mechanism can be “universal”’. In a lecture delivered at the University of Illinois in 1949 (entitled ‘Rigorous Theories of Control and Information’) von Neumann said:
The importance of Turing’s research is just this: that if you construct an automaton right, then any additional requirements about the automaton can be handled by sufficiently elaborate instructions … [A]n automaton of this complexity can, when given suitable instructions, do anything that can be done by automata at all.
The Los Alamos physicist Stanley Frankel, responsible with von Neumann and others for mechanizing the large-scale calculations involved in the design of the atomic and hydrogen bombs, has described von Neumann’s attitude to Turing’s work:
I know that in or about 1943 or ’44 von Neumann was well aware of the fundamental importance of Turing’s paper of 1936 ‘On computable numbers …’, which describes in principle the ‘Universal Computer’ of which every modern computer (perhaps not ENIAC as first completed but certainly all later ones) is a realization. Von Neumann introduced me
to that paper and at his urging I studied it with care. Many people have acclaimed von Neumann as the ‘father of the computer’ (in a modern sense of the term) but I am sure that he would never have made that mistake himself. He might well be called the midwife, perhaps, but he firmly emphasized to me, and to others I am sure, that the fundamental conception is owing to Turing – insofar as not anticipated by Babbage, Lovelace, and others. In my view von Neumann’s essential role was in making the world aware of these fundamental concepts introduced by Turing and of the development work carried out in the Moore school and elsewhere.
Given our knowledge of the achievements at Dollis Hill and Bletchley Park, the history of computing must be rewritten. Histories written in ignorance of Colossus are not only incomplete, but give a distorted picture of the emergence and development of the idea of the modern computer. Turing’s logical work in 1935–6 and Flowers’ work at Bletchley led, via Newman’s desire to put the concept of the stored-program universal computing machine into practice, to the Manchester Computing Machine Laboratory and the Manchester Mark I computer. From this in turn came another momentous development, the first mass-produced computer to go on sale, a copy of the Mark I. In the US, Turing’s work steered von Neumann to the underlying logical principles of the EDVAC and the IAS computer. Technology transferred from Manchester, the Williams tube random-access memory, was crucial to both von Neumann’s IAS machine and IBM’s first mass-produced stored-program computer, the IBM 701. The 701 was a foretaste of the global transformation soon to flow from this criss-crossing pattern of invention and influence set in motion by Alan Turing in 1936.
Chapter 20 reveals what the Germans knew about Enigma's security. Plugboard Enigma would have been impregnable if it had been used properly, but with up to about 40,000 Enigmas in service there was little prospect of that happening - the fundamental mistake made by the
Wehrmacht
was to think otherwise. However, it should never be forgotten that plugboard Enigma could be broken regularly by âpure' cryptanalysis only in one set of circumstances - when it used doubly enciphered message keys. After April 1940, only the
Kriegsmarine
made the catastrophic mistake of employing that very fallible indicating system, with no fewer than six ciphers. All other plugboard Enigma ciphers could be broken only with a good crib, which had to occur on a daily basis before a cipher would yield to
the bombes. Since Hut 6 was attacking over sixty
Heer
and
Luftwaffe
Enigma ciphers at any one time later in the war, the
Germans had continually to make a lot of mistakes every day for it to succeed against them. If the
Heer
and
Luftwaffe
had monitored their radio nets more carefully, and given the OKW's radio security organization some real power at an early date, GC&CS would have had many more problems in breaking Enigma - and Hut 6 Ultra would have been thin on the ground. As it was, introducing even the most basic precautions in preparing messages for encipherment made Hut 6's task much more difficult, and sometimes an impossible one: one such change quintupled the amount of bombe time required to break daily keys.
The
Wehrmacht
paid dearly for its decision to make Enigma a standard machine, common to all three of its branches, with only the
Kriegsmarine
employing extra rotors. The British did not make the same mistake with their own cipher machine, Typex, even though it was essentially a copy of commercial Enigma. Perhaps profiting from their experiences in breaking Enigma, there were about twenty different sets of rotors for Typex, which made the Germans give up their attempts to break it (and helped to protect it against US Navy attempts to solve it in 1945). Nor did the Americans overlook the principle that making good codes and ciphers is as important as breaking them. Not only did the United States produce a large number of different rotor sets for the Navy's very advanced Electric Cipher Machine (the Army's Sigaba), but special codebreaking units were tasked with attacking their own cryptographic systems, both in theoretical studies and, more importantly, as they were used in practice. One machine was found to be wanting, but special steps were taken to increase its security.
Chapter 20 sets out the findings of some of the
Kriegsmarine
's many inquiries into Enigma's security. It also describes devices and precautions introduced by the
Wehrmacht
to improve Enigma, including one which would have had devastating effects for Ultra - if only it had been used properly. As Gordon Welchman observed, the Allies were indeed lucky.
RE
There has long been uncertainty about what the Germans knew about Enigma's security. Fortunately, the release of some important post-war
TICOM (Target Intelligence Committee) interrogation reports of German Sigint personnel has at last thrown considerable light on this issue. Towards the end of the war, TICOM made plans to round up the staff from the numerous German Sigint agencies, in order to find out exactly what they knew about German cipher machines and code-breaking equipment, and the extent to which the Germans had been able to penetrate Allied codes and ciphers. When the war ended, and for some while afterwards, TICOM units combed Germany for Sigint personnel and equipment and made many valuable finds.
The German Foreign Office refused to use Enigma, since it was not convinced that Enigma was secure. Instead, it employed one-time pads to encipher its highest level traffic, in a system known to the Allies as GEE. However, GEE itself was insecure: the pads being used did not contain fully random numbers, since they had been generated by a machine which produced predictable sequences. Cryptanalysts in the US Army's Signal Security Agency exploited this to solve GEE in the winter of 1944â5.
A
Heer
or
Luftwaffe
Enigma key-list comprised the rotor order
(Walzenlage)
, ring settings
(Ringstellungen)
and plugboard settings
(Steckerverbindungen)
. A naval list also set out the
Grundstellungen
. Cryptanalysts attacking Enigma were therefore confronted by a huge key space: 60 (5Ã4Ã3) rotor orders, 676 (26Ã26) effective ring settings, 150 million million
Stecker
combinations, and 17,576 (26Ã26Ã26) message starting positions â a combined total of 1.074Ã10
23
. The total was even higher for the standard naval machine M3 (6.0144Ã 1023), and twenty-six times greater again for M4 â 104 times after 1 July 1943, when a second âGreek' rotor (gamma) and its related thin reflector were introduced.
These are huge numbers, but some experts in the
Kriegsmarine
were not impressed by them. Soon after the beginning of the war, German naval cryptanalysts advised that Enigma was not secure. A report prepared by Wilhelm Tranow, then a senior member of the
Kriegsmarine's
codebreaking agency, the
B-Dienst
, recommended that the
Kriegsmarine
should replace Enigma by a codebook. At first sight that appears to be an odd recommendation since the
B-Dienst
was already breaking several British enciphered naval codes, and Tranow was fully aware of the weaknesses of such codes generally. However, Germany's circumstances were very different from those of Britain, since its Navy did not have to operate on a global basis.
Tranow's recommendation might just have worked, particularly if the enciphering additive tables had not been overused. Once a daily Enigma key was found, all the messages in it fell, but each message in an enciphered code has to be solved separately, and it is a very slow process until the additive book being used can be substantially reconstructed.
The
Kriegsmarine
, in particular, carried out many inquiries into Enigma's security. Captain Ludwig Stummel, of the Marine Communications Service, the
Marinenachrichtendienst
(MND), reported on it following the capture of
Schiff
26 in April 1940, and the loss of U-13 off the coast of Norfolk on 30 May 1940. On both occasions he gave assurances that Enigma was safe to use: all the important documents relating to it were printed on specially absorbent water paper, using water-soluble ink so that they should have been destroyed if a U-boat was sunk. And even if key-lists had been captured, a special
Stichwort
(cue word) procedure, which modified the settings, had been brought into force after the captures. But Admiral Karl Dönitz, the Admiral Commanding U-boats
(Befehlshaber der U-Boote - BdU)
remained deeply uneasy. Following the ambush of U-67 and U-111 by the British submarine HMS
Clyde
in Tarafal Bay in one of the Cape Verde Islands, off the coast of Senegal, on 28 September 1941, he concluded that:
Either our ciphers have been compromised or it is a case of leakage. A British submarine does not appear by chance in such a remote part of the ocean. The Naval Staff is requested to take the necessary measures to safeguard the cipher system.
In the ensuing inquiry, Rear-Admiral Erhard Maertens, the head of the MND, advised that âThe more important ciphers do not seem to have been compromised.'
As the war progressed, the evasive routeing of Allied convoys, attacks on U-boat rendezvous and evidence from
B-Dienst
decrypts, led the
Kriegsmarine
, especially its U-boat arm, to become increasingly alarmed about cipher security. The
B-Dienst
was solving many of the âU-boat estimates' promulgated by the British and Americans. Although these were based on Ultra in 1943, they invariably referred to the intelligence being derived from direction-finding (DF), to disguise its true origin. In late March 1943, for example, BdU recorded that:
according to an American estimate of the U-boat situation on 24 March it is presumed on the basis of inaccurate D/F that 10 to 15 U-boats are thought to be patrolling in the area 56 to 61 degs. N and 28 to 36 degs. Wâ¦. The Seeteufel Group with 14 boats was in a patrol line from AD 7986 to AK 3955 up to 2100 on 23 March.
However, as BdU observed, DF could not account for the estimate, since the
Seeteufel
group had not yet sent any signals. Even so, BdU managed to convince himself that the American estimate could have been deduced from a detailed analysis of all the available intelligence, and that it was not the result of the Allies reading the U-boat signals.
BdU's suspicions were again aroused by a series of attacks on U-boats at or near no fewer than ten refuelling rendezvous in mid-Atlantic between 3 and 11 August 1943. These were followed by a highly disturbing report from the
Abwehr
in Switzerland:
A special office [in England] has dealt exclusively with solving German codes. It has succeeded for some months in reading all orders sent by the Kriegsmarine to U-boat commanders, which has very considerably helped the hunt against the U-boats.
The
Abwehr
had received the information from a member of the Swiss intelligence service. The original source was said to be a SwissâAmerican in an important position in the US Navy Department. BdU's alarm can be imagined, but the experts reaffirmed their earlier advice that the âcontinuous current reading of our W/T traffic by the enemy is out of the question', since âAt present no possible way of solving [the plugboard combinations] within a reasonable time is known, even with the maximum amount of labour.'
At the end of February 1944, the MND carried out a further inquiry into how the Allies had found out about three rendezvous, between U-129, U-516 and U-544 on 16 January, the tanker
Charlotte Schliemann
and U-532 on 11 February, and U-518 and the Japanese submarine I-29 on 13 February. However, the committee conducting the inquiry does not seem to have approached its task with an open mind, since it explained its purpose as being âto explain for what reasons reading of our signals ⦠could
not
have taken place' (my emphasis). It completely discounted the possibility that messages had been read with the help of punched card systems, since, for
some unexplained reason, it thought the systems âwould have to be applied anew to each signal'. A later investigator, Lt. Hans-Joachim Frowein, did not make the same mistake. The committee concluded that if âthe enemy really was for a limited time in a position to read our traffic currently it is incomprehensible that during this period he should not have interfered with all U-boat operations'. The fact that doing so would have endangered the Allied source seems not to have occurred to them. Since reading the traffic was âshown to be out of the question', only the usual culprits â âtreachery or discovery by enemy aerial reconnaissance' â could explain the incidents. As it was, the report had scarcely been completed before U-It 22, an ex-Italian freight-carrying U-boat, was sunk on 11 March at a rendezvous some hundreds of miles south of Capetown, and a refuelling tender,
Brake
, was attacked and destroyed by a British destroyer on 12 March, the day after a rendezvous in the Indian Ocean. This immediately led to yet another inquiry, and later to a major report on
Kriegsmarine
cipher security generally.
The
Kriegsmarine's
investigators continually emphasized the security afforded by the
Stichwort
procedure in protecting key-lists which might have been compromised by capture, since it was obvious that the Allies would use a captured key-list and then try to find message settings by trial and error. The
Kriegsmarine
even believed that the
Stichwort
procedure protected captured key-lists from attacks using âextensive mechanization'. But the investigators were completely mistaken about the
Stichwort's
effectiveness, and can scarcely have given it even a cursory analysis. Under the procedure, five numbers were derived from the letters in a secret key-word
(Kennwort)
, for example â
baden
' furnished 2, 1, 4, 5, 14. The first number was added to each of the rotors which would otherwise have been used (with 2, rotors III, I, VII became V, III, I), the next three to the
Ringstellung
, and the last number to each
Stecker
letter. An attacking cryptanalyst therefore only had to try a maximum of eight new rotor orders (instead of 336) and a mere twenty-six sets of
Stecker
(instead of 150 million million), which could be done even by hand methods, and would have taken only a few minutes with punched cards. It is amazing that the
Kriegsmarine
placed so much emphasis on such a weak procedure, demonstrating as it does that the people carrying out the inquiries had no cryptographic expertise. Even more surprisingly, it would seem that they did not ask any real experts to examine the procedure carefully.