Wednesday, May 7, 2014

Compromise of the State Department’s strip cipher in 1944

During WWII the US State Department used several cryptosystems in order to protect its radio communications from the Axis powers. For low level messages the unenciphered Gray and Brown codebooks were used.  For important messages four different codebooks (A1,B1,C1,D1) enciphered with substitution tables were available. Their most modern and (in theory) secure system was the M-138-A strip cipher. Unfortunately for the Americans this system was compromised and diplomatic messages were read by the Germans, Finns, Japanese, Italians and Hungarians. The strip cipher carried the most important diplomatic traffic of the United States (at least until late 1944) and by reading these messages the Axis powers gained insights into global US policy.

The strip cipher was not a weak system cryptologically, even though it could not offer the security of cipher machines. The success of German and Finnish codebreakers was facilitated in many cases by the poor way that the system was used by the State Department.
M-138-A strip cipher

The M-138-A system consisted of an aluminum frame (or later wooden/plastic) with room for 25 or 30 paper strips. Each strip had a random alphabet. The daily key specified the strips to be inserted and the order that they were to be inserted in. The plaintext was written vertically at the first column by rearranging the strips. Then another column was selected to provide the ciphertext.
Each embassy or consulate had 100 alphabet strips, 50 ‘circular’ alphabet strips and 50 ‘specials’. The ‘circulars’ were used for communications between embassies and for messages from Washington to several embassies. The ‘specials’ were used for direct communications between Washington and a specific embassy.

The way the system worked was that each day 30 alphabet strips were chosen out of the available 50 (both for the ‘circulars’ and the ‘specials’). The strips used and the order that they were inserted in the metal frame was the ‘daily key’. The strip system did not have a separate ‘key’ for each day. Instead there were only 40 different rearrangements.
German efforts to solve the US diplomatic strip cipher

Three different agencies worked on the US diplomatic M-138-A strip cipher. The German High Command’s deciphering department – OKW/Chi, the Foreign Ministry’s deciphering deparment Pers Z and the Air Ministry’s Research Department - Reichsluftfahrtministerium Forschungsamt.
At the Forschungsamt some work was done on the strip but apart from the fact that they solved some traffic we don’t know any more details. 

At OKW/Chi an entire team worked on the strip, led by the mathematician Wolfgang Franz and they built a specialized cryptanalytic device called ‘Tower clock’ (Turmuhr). This device was a ‘statistical depth-increaser’ according to US reports.
At  Pers Z they devoted significant resources against the strip cipher. A team of mathematicians, led by Professor Hans Rohrbach made extensive use of IBM/Hollerith punch card equipment in their efforts to solve the alphabet strips and also built a special decoding device called ‘Automaton’.

Proof of OKW/Chi success in 1944
The information given by Wolfgang Franz who was interrogated in 1949 is limited. In his report DF-176 he said in pages 6-9:

‘Especially laborious and difficult work was connected with an American system which, judging by all indications was of great importance. This was the strip cipher system of the American diplomatic service which was subsequently solved in part.’
All told, some 28 circuits were solved at the Bureau under my guidance, likewise six numerical keys-some of them only in part.’

A matter of some controversy is the extent of success they had in 1944 against this system.  The head of the mathematical research department of OKW/Chi, Dr Erich Huettenhain said iTICOM I-2 ‘Interrogation of Dr. Huettenhain and Dr. Fricke at Flensburg, 21 May 1945’, p2:
‘Q. What work was done on British and American codes and ciphers?
A. Diplomatic - most of the American strip cipher was read, strip cipher was used by the military as well as by the diplomatic.’

However in TICOM I-145 ‘Report on the US strip system by Reg Rat Dr Huettenhain’ he stated:
‘Only a little of the material received could be read at once. Generally it was back traffic that was read. As, however, the different sets of strips were used at different times by other stations, it was possible, in isolated cases, to read one or the other of the special traffics currently. We are of opinion that of the total material received, at the most one fifth was read, inclusive of back traffic. None was read after the beginning of 1944.’

This seems to be at odds with the version given by the same person in an unpublished manuscript written in 1970 in which he said: 
‘Auf diese Weise wurden von 1942 bis September 1944 insgesamt 22 verschiedene Linien und alle cq-Sprüche mitgelesen’

TranslationIn this way, were read by 1942 to September 1944, a total of 22 different links and all cq (call to quarters) messages. (note that cq messages means ‘circulars’)
Were the Germans able to solve the State Department’s high level messages in 1944? The answer is yes.

In the US National Archives, in collection RG 457 ‘Records of the National Security Agency’ - Entry 9032 - boxes 205-213 ‘German decrypts of US diplomatic messages 1944’ one can find many decoded messages from US embassies and consulates around the world. Many have a note on the lower right side identifying the cryptosystem used. The German code for the strip cipher was Am10. This is mentioned in TICOM I-145 which says ‘The American strip system Am10’ and in TICOM DF-176, p7: ‘the Am10-that was the designation of the strip cipher system’.
In these boxes there are a few messages with the tag Am10 sent in 1944 and decoded in that year. They prove that the Germans could solve the strip system even in 1944. Here are four of these messages:

From box 209 – Bern-London

From box 210 – Madrid-Algiers

From box 209 - Algiers


From box 212 – Madrid-Washington

Messages between embassies should have been on the ‘circular’ strips. Messages to or from Washington should have been sent on the ‘special’ strips. From the TICOM reports and the few messages found in boxes 205-213 it is clear that the German codebreakers were able to solve the strip cipher even as late as 1944 and that included both the ‘circular’ messages and at least some of the ‘specials’.

In addition there is in these boxes a list with the code L-1456 vol VIII that according to NARA ‘does not appear to be linked to the other documents’. It is possible that it has some connection to the M-138-A case.

Acknowledgements: I have to thank Randy Rezabek of TICOM Archive for collaborating with me on this research project and covering parts of the cost and also my researcher Mike Constandy of Westmorland Research for going though the boxes and finding needles in a haystack.

No comments:

Post a Comment