How the Germans broke US Code B7 – Part Two

well_chuffed, Going Postal

The previous part described what a two part code is and what you need to be able to break it. This is Dr Müller’s description of how the Germans broke the American B7 code.

The first part involved creating the lists of groups detailing how and where they were used. There were three bookbreakers who had a number of assistants. Each day had 2 shifts , the first tried to work out the meaning of the groups , the second indexed the results of the first shift. They still worked on further processing crating special lists. These contained all passages in which the same group occurred and the laying out of a “Sach-Code” , that is a sort of encode arranged by subjects.

The material they worked on was received from the workshops of the cryptanalysts and a the previously referenced Original , Material I and Material II. Original contained the enciphered text and was useful in case there was any garble.

As well as the special lists , other work was also done on the messages. Numerous frequent groups had been details in special lists , repeats in the tests had been underscored and put into further lists and a circuit catalog had ben prepared.

The bookbreakers studied this material closely and on the second day got their first break. Two of the most frequent groups , BIBOT and KUJAN were suspected of being PERIOD or PARAGRAPH since their distribution through the text suggested the occurrences of a period in the language. This interpretation was facilitated by the fact that numerous telegrams had “Section 1” , “Section 2” , “Section 3” etc. in clear text as prefixes. The individual sections often began or ended with BIBOT or KUJAN.

Two other groups had also attracted attention , these were DODEV and GYBUX which always appeared as an associated pair and were separated by a variable number of groups. The assumption was made that they were OPEN QUOTES/CLOSE QUOTES or OPEN/PARENTHESIS/CLOSE PARENTHESIS or dashes between which something was enclosed.

These interpretations did not give away any of the message content but did give valuable clues about the formal construction of a message.

The actual break-in usually results from message beginnings. Generally telegrams begin with reference to an earlier message , fo example “My telegram number 267 of 6 February 10 o’clock a.m.” or similar. In the case of B7 the beginnings were not favourable. Frequent groups did occur at the beginning which probably meant “my telegram” , “your telegram” , “department’s telegram” etc. but the following groups that should contain the number and date did not show the repetitions to be expected. The Americans had the habit of sending messages on the same circuit using different code systems. Intervening numbers were in another system. The numbers were represented by a single group rather than a group for each digit and the same for the dates , there were 366 different groups for dates.

The first break came from the end of the message rather than the start. Often toward the end of the telegram there was a PERIOD (BIBOT or KUJAN) followed by only a few groups , sometimes only two. The group after the period was often GUDOR. This occurred almost always messages that were sent simultaneously to two or more stations. Reference to this was often made at the beginning with the words “Following telegram sent to department” not in the code of the message but in another , the B3 or Grey Code which had ben solved years ago. Figuring out where the message was sent , say Washington and Moscow , the message might end with the words “Repeated to Moscow”. At then end of this telegram were three groups – BIBOT (PERIOD) GUDOR KOMYB. This means GUDOR must be REPEATED TO” and KOMYB “MOSCOW” or “REPEATED” and “TO MOSCOW”. If GUDOR meant “REPEATED TO” it could never be at the end of a message and it never was.

The next step was to study all messages with GUDOR near the end and determine to which stations they had been sent. In this the codes for a whole series of city names was established. Those set to Ankara yielded really favourable results , there were always three groups at the end. Code B7 was very old , from about 1920. At that time Ankara was not the capital of Turkey , that happened in 1923 and in fact the Ottoman Empire was being dismantled with the prospect of a rump of Anatolia forming the new Turkey.

Ankara message had three groups after GUDOR but with variations


The middle group was always the same , the first and last parts had variations. Parts 1 and 3 were frequent groups that appeared in many place , DAKUC was very rare. They were three components of AN-KAR-A. AN and A are the indefinite article.

This was a serious advance , not only did they have A and AN but they then indicated if the following group began with a vowel (or silent H) or a consonant. To top it all , they had the first letter of the alphabet.

They had similar luck with Istanbul which broke down into IST-AN-BUL. IST being a frequent suffix such as economist , nationalist or socialist. The group for BUL hardly ever appeared outside this combination. The group for AN included those used for the AN in ANKARA.

The next task was to find the prepositions used before the city names. You would expect to find IN , FROM , TO etc. For those groups where the city name was a group only few previous groups were found repeated. For Ankara , Istanbul and Vichy there were numerous repetitions. This suggested that “IN LONDON” , “FROM PARIS” , “TO BERLIN” had their own groups whereas for spelled names the encoder had a provide a separate group for the preposition.

The progress of the solution went along these lines. The preparation work of preparing and extracting the groups must have taken a long long time.

© well_chuffed 2018

Audio file