The titans who won key battle

Ralph Erskine cracks open a tale of the men and their Heath Robinson machines that decrypted German ciphers

Colossus was the special-purpose electronic computer used by Bletchley Park code-breakers to solve messages enciphered on the German Lorenz SZ 40/42 teleprinter cipher machine, which was code-named Tunny.

Colossus entered service in February 1944, but Bletchley had attacked Tunny manually from 1941 onwards. A substantial part of this book explains how Tunny messages were broken, while about ten of its 48 chapters and appendices concern Colossus itself.

Cracking Tunny traffic was initially slow. Later, "depths" (messages transmitted with identical machine settings) and poor German procedures enabled Bletchley to read nearly every Tunny message from June until October 1942, when a weak indicating system was changed. Tommy Flowers, the inventor of Colossus, describes being astonished that "the Germans never realised that their secret code could be broken". However, wartime German records and a postwar US Target Intelligence Committee (Ticom) interrogation report show that they knew about Tunny's weaknesses, which is why they improved it constantly.

In a brilliant feat of cryptanalysis, Bill Tutte, a diffident Cambridge University chemistry research student, had deduced Tunny's structure from a stretch of Tunny key (characters used to encipher plain text) found by Colonel John Tiltman in August 1941. Bletchley's research section then joined in to solve the complete machine by January 1942. They discovered that Tunny incorporated two sets of five mechanical wheels, with one set stepping irregularly, controlled by two so-called motor wheels. The wheels had a total of 501 settable cams that produced the equivalents of teleprinter "marks" and "spaces" and endowed Tunny with an astronomical key-space. The output of both sets of wheels was combined to generate streams of key, which were added to Baudot-Murray teleprinter code to encipher it.

As Tunny became more complex, breaking messages manually became increasingly difficult. However, being electromechanical, Tunny could produce only pseudo-random key, which left various statistical and linguistic characteristics in messages open to exploitation. Tutte invented a statistical method for breaking single Tunny messages in November 1942, but it would have taken years to manually make the calculations for even one message. Max Newman, a Cambridge mathematician (and a "supreme facilitator") proposed fast machinery for the purpose. The device, nicknamed Heath Robinson after the cartoonist, entered service in June 1943. It had few valves and was not wholly successful, being slow and unable to fully synchronise the two paper tapes being compared. But it was vital to subsequent successes because it revealed problems that were later eliminated when designing Colossus.

In a parallel development, and in spite of Bletchley's scepticism, Flowers and a gifted Post Office team had slaved on Colossus for almost a year without any official requisition. Colossus I ran a looped tape containing cipher text at 30mph, while about 1,500 valves, an unprecedented number, performed tests and emulated Tunny functions. Colossus I counted the number of times certain events occurred and recorded scores - with the highest revealing some wheel settings. Like later Colossi, it lacked stored programming. Fortunately for Flowers, both his boss, Gordon Radley, and Sir Stanley Angwin, the Post Office chief engineer, had backed him by making considerable design and production facilities available. Otherwise, Colossus II with its 2,500 valves, would not have been in service before D-Day to provide crucial intelligence about Hitler's reactions to Allied deception plans. Colossus II used parallel processing to handle data at 25,000 characters a second - a speed that postwar computers did not reach for some years.

Flowers felt that he did not receive the recognition he was due, and he was embittered. He was overlooked in a 1947 Post Office reorganisation, and he thought that his proposals for postwar electronic telephone exchanges were ignored. But one wonders: electronic telephone exchanges proved very difficult to develop, and the 1963 Highgate Wood trial exchange was "a magnificent failure".

Tunny, which was used mainly between the German High Command and army groups, provided more strategic intelligence than Enigma. Not many Tunny decrypts survive; so little Tunny intelligence appears in this book.

However, although not mentioned here, Tunny decrypts between June and October 1942 revealed that Kriegsmarine B-Dienst code-breakers were solving the principal Allied convoy code. In a massive blunder, that code was not replaced until June 1943, resulting in huge losses of ships and human life, especially during the deadly convoy battles of March 1943. The delay was largely organisational, but senior managers at Bletchley cannot have pressed hard enough to get it changed.

Breaking Tunny traffic was the greatest code-breaking feat of the war. The US Army's reconstruction of Japan's Purple diplomatic cipher machine was comparable to Bletchley's solution of Tunny, but ascertaining Purple's daily settings was relatively simple; whereas finding Tunny's wheel patterns and settings required the highest cryptanalytical skills and involved advanced statistical techniques and some of the most complex electronic equipment of the war. Even intercepting Tunny signals posed immense problems; it eventually required more than 800 staff at Knockholt listening station near Sevenoaks. From November 1942 onwards, Knockholt intercepted almost 168,000 transmissions, which yielded 13,500 decrypts, with 63 million characters; more than 70 per cent of the decrypts were made after June 1944, when three or more Colossi were in service.

Colossus includes contributions by the machine's designers and the cryptanalysts (such as Tutte and Donald Michie) and operators who used it, as well as intelligence historians. It should put paid to myths that are still so prevalent that they appear on a national museum's website: that Alan Turing "helped to invent" Colossus (he did not, although he made major contributions to methods for solving Tunny messages), and that it "cracked the Enigma codes".

Colossus is inevitably quite complex at times (for examples of breaking Tunny, visit www.codesandciphers.org.uk/anoraks/pods ); but with application, non-mathematicians should be able to follow most of it. Even in the technical appendices, Jack Copeland and Frank Carter explain with admirable clarity the complexities of "Turingery" and "rectangling". Some chapters are versions of papers that have appeared previously, sometimes in obscure places. Most achieve a high standard, although a few should have been omitted, and others appear to have been cut too heavily. This book will appeal to anyone interested in Colossus, code-breaking or Bletchley Park. Copeland and the other contributors have rightly done Flowers and the Tunny code-breakers proud.

Ralph Erskine is a retired barrister who has written extensively on signals intelligence.