Virtual Colossus

The first break

The first intercepts, the Tiltman break and Bill Tutte’s contribution.

The first intercepts

Early in 1940, a group of policemen on the south coast were listening out for possible German spy transmissions from inside the UK, but instead they found enciphered teleprinter signals being transmitted.

Following this discovery, a Y-station was set up at Knockholt in Kent to intercept and record this traffic. The operator used an undulator to record the impulses of the signal onto a narrow paper tape called a slip.

A strip of undulator slip, showing the signal's impulses as a line with raised and lowered sections

The slip shows the impulses and gaps. These were carefully read by hand, punched onto paper tape and sent to Bletchley Park.

The Tiltman break

Brigadier John Tiltman, one of the top codebreakers at Bletchley Park, became interested in these messages, especially ones where the German operators made mistakes and sent more than one message with the same start positions (known as a depth). Each message (in the first part of the war) was sent with a 12-letter indicator at the start, which told the receiving operator how to set up his Lorenz machine. If two messages had the same indicator, a depth had possibly occurred, and these were sent to Bletchley Park. Not much headway was made into breaking the cipher until the Germans made one horrendous mistake.

On 30 August 1941, two messages of nearly 4,000 characters were intercepted with the same 12-letter indicator at the beginning, HQIBPEXEZMUG. It appears the first operator had set up his machine as normal and typed in nearly 4,000 characters of message by hand, which was transmitted. Unfortunately for them, what came back from the receiving end was the German equivalent of “sorry, didn’t get that, send it again”.

It was then that the mistake was made. Rather than using a new setting, the operator set his Lorenz machine back to the same start position and began typing the message again. This was absolutely forbidden, but they did exactly that. If he had typed exactly the same message, word for word, then exactly the same cipher text would have been transmitted the second time and Bletchley Park would have gained nothing.

Understandably annoyed, probably, at having to type out all that message again, the operator started to shorten parts of it. For example, he began the first message with the German word SPRUCHNUMMER, which means “message number”, but in the second he keyed SPRUCHNR and then carried on. A person can read “message number” just as easily as “message no.”, so what difference does it make? From just after the N the two texts were different, but both were enciphered with the same obscuring key.

Using some clever detective work, and knowing that the cipher text was produced with a teleprinter and the Vernam cipher, John Tiltman was able to separate out the key that had been used to encipher the text.

This is how it works. Remember, Lorenz enciphers the text like this:

Plain + Key = Cipher text

So if we add the two cipher texts together, where the key is the same, we get:

Cipher text 1 + Cipher text 2 = Plain 1 + Plain 2 + Key 1 + Key 2

As we know these two messages are in depth and used the same key, the two keys cancel out, leaving just the two plain text messages mixed together:

Cipher text (1 + 2) = Plain text (1 + 2)

Separating the two messages

Once he had added the two messages together, breaking apart the two plain texts was a bit like solving a logic puzzle.

The first thing to do was think of a word or two that was very likely to be in the message. At Bletchley Park it was known that the German army numbered its units in Roman numerals, so German words such as “ROEM” (Roman) followed by a number such as “EINS” (one) could be expected in messages. Another frequent word was “GEHEIM”, meaning secret.

This word was then dragged through the combined Plain 1 + Plain 2 text, adding it to one section at a time. If the result was gibberish, it was in the wrong place and was tried somewhere else.

Example

Cipher (1 + 2)HCVFODJSP9/
Check text9ROEM9EINS9
ResultT9IN9FRAGE9

In the example above, adding the text 9ROEM9EINS9 (9 was how a space was written) gives the words “IN FRAGE”. This could then be guessed as part of the phrase “KOMMT NICHT IN FRAGE”, German for “is out of the question”. The fuller text was then added back to the combined cipher text to try to extend it further.

In this way Tiltman managed to tease out both messages. By then the messages themselves were out of date and no longer useful, but it meant that, using both the cipher text and the now-known plain text, he could work out the key used to encipher it:

Cipher text + Plain text = Key

Tiltman now had a stretch of roughly 4,000 characters of key generated by the enciphering machine. They still had no idea exactly what was generating it, but it was a great breakthrough.

Bill Tutte’s contribution

After a number of codebreakers had failed to make any further headway, the messages and key were given to a new codebreaker who had recently joined Bletchley Park: Bill Tutte.

Having recently completed a course at cryptographic school, Bill Tutte had learned that one way to break a cipher was to look for repetitions in the key by writing it out in tables.

After months of trying, he wrote out just the first impulse of the key in rows of 41 and began to see patterns emerge, which meant that a repetition period of 41 had some significance in the way the key was generated. This turned out to be the number of positions on the first Chi wheel.

As it wasn’t a perfect repetition, he also deduced that two wheels, rather than one, were generating the key. He called these the Psi and Chi wheels. He also worked out that one set of wheels stepped on every letter while the other stuttered, and from this deduced the motor wheels needed to do it.

Over the next couple of months, Tutte and the other members of the Research Section worked out the complete logical structure of the cipher machine we now know as Lorenz. They did not see an actual Lorenz machine until right at the end of the war, which finally confirmed that it was built in exactly the way they had worked out.

After this huge intellectual feat, he went on to find a way of statistically working out the start positions in smaller pieces, using just two wheels at a time. But this meant checking 1,271 possible start positions for just the first two wheels: a huge amount of work to do by hand, by which time the messages were too out of date to be useful. What was needed was a mechanised way of running this calculation: a machine to beat a machine.

More details of this statistical attack are on the next page, the statistical attack on Lorenz.