Tuesday, August 21, 2012

Soviet cipher teleprinters of WWII

At the start of the Soviet-German war the cipher machines used by Soviet authorities were the K-37 ‘Crystal’ (a modified Hagelin B-211) and the cipher teleprinters B-4 and M-100.

Few details about these machines are known. I have covered the K-37 here. An article in agentura.ru says that the B-4 was built in the 1930’s and used in Spain during the civil war, in the Far East border incidents with Japan and in the Russo-Finnish war of 1940.

A modernized version of the B-4 called M-100 was built in 1940 and installed in American busses in order to provide mobility (it weighed 141kg). By the summer of 1941 there were 96 sets of the M-100 in service.

In 1943 a more compact version of the M-100 was built and given the designation M-101.

The agentura piece does not give details on the characteristics of these machines.

German interception of Soviet radio-teletype networks

During the 1930’s the SU started to use radio-teletype. Since 1936 the Germans had equipment that automatically intercepted and printed this traffic.

From TICOM reports it seems that the military networks used 2-channel teleprinters while the economic links used multichannel systems.

The 2-channel links employing cipher T/P assigned one channel for the cipher traffic and the other for operator ‘chat’.

There were at least three agencies that intercepted this traffic.

1). The Army Ordnance, Development and Testing Group, Signal Branch Group IV C  -  Wa Pruef 7/IV C at an experimental station in Hillersleben-Staats.

2). The Forschungsamt, an agency reporting directly to Hermann Goering.

3). The Army’s signal intelligence agency OKH/GdNA (General der Nachrichtenaufklaerung) Group VI - OKH/GdNA Group VI.

Forschungsamt success in 1943

In 1943 the Forschungsamt informed the Army agency of their success in reconstructing a Soviet cipher teleprinter used between Moscow and the Fronts. This machine was solved because during every pause seven characters of pure ‘key’ were transmitted. A meeting was held in September 1943 between Councilor Paetzel (head of the FA’s cipher research department), Councilor Kroeger (the FA’s cipher machine specialist) and the Army agency’s Dr Pietsch (head of the mathematical research department) and Doering (head of cipher machine research).

The Soviet machine had 6 wheels. Five enciphered the respective Baudot impulses while the sixth controlled their movement.

The wheels had the following positions (pins):

wheel I – 49

wheel II – 47

wheel III – 46

wheel IV – 45

wheel V – 41

wheel VI – 43

The Soviet scrambler corresponded to the left portion of the German SZ 40/42 cipher attachment. The Forschungsamt people stated that they would build a copy of this machine in order to decrypt this traffic more efficiently.

Unfortunately we don’t know any more details about this affair.

Efforts of the Army signal intelligence agency

The army agency assigned its own unit to intercept and evaluate this traffic. This was Group VI operating during 1942-44 from Loetzen, East Prussia.

Unteroffizier Karrenberg, the member of that unit assigned to work on the cipher teleprinter said in postwar interrogations that this traffic was first intercepted in 1940 in Warsaw. However it was not systematically collected and analyzed till summer 1943. They called this traffic ‘Bandwurm’ because of the non-repeating cipher.

There were 8 T/P links from the Army Fronts to Moscow plus 2-3 Airforce links and a link to the Far East command. It was also used by the NKVD. There was no direct T/P link between the individual Front staffs. Instead messages had to be routed through Moscow.

According to Karrenberg the machine had two settings a large and a small. The large setting gave a simple substitution because the wheels did not turn. This was used for operator ‘chat’. The small setting gave an endless column substitution since the wheels moved.

In TICOM report I-153 he says: ‘In Autumn 1944 both the end of 'adder' and every pause in the cipher proper was preceded by seven key letters [redacted]. Then the traffic went off the air and reappeared in December with no external change except that the seven ‘residue' letters had been reduced to three, suggesting a modification of the machine’. In TICOM report I-30 he says that the attachment had 5 small wheels driven by a large one with a period of 43.

These were the same characteristic observed in the machine analyzed by the Forschungsamt. However another OKH cryptanalyst named Buggisch says that the FA machine and the ‘Bandwurm’ were different.

Buggisch was assistant to Doering in the cipher machine section of OKH/GdNA. In his interrogation TICOM I-64 he says that the cycle of one wheel was 37 and the others 30-80. The machine was analyzed by the mathematics department and a cryptanalyst Troeblicher or Troebliger played a leading part. Thanks to a ‘compromise’ of 8 messages enciphered with the same settings 1.400 characters of pure ‘key’ were recovered. However they were not able to solve the machine because they lacked the manpower.

Unfortunately Buggisch left the OKH agency in June 1944 so he did not know anything more.

The Germans may have failed to solve this machine but they were able to decode messages ‘in depth’ by anagramming. A machine was built that automatically printed the Baudot traffic in Hollerith/IBM cards and these were searched for repeats but with limited success.

The intercepted messages contained reports on Soviet and German military dispositions, , statements by POW's, signal intelligence reports, reports for TASS and SOVINFORMBUREAU, letters concerning postings, transfers, promotions, weather situation reports and supply manifests.

The Poets series cipher machines

In the immediate postwar period the Anglo-Americans were able to solve at least two Soviet cipher teleprinters. These were given the names Coleridge and Longfellow.

Coleridge was used on military networks in the European part of the Soviet Union. By March 1946 it had been solved. The Coleridge decrypts provided important intelligence about the Soviet military’s order of battle, training activities and logistical matters.

The other system, Longfellow, was reconstructed by July 1946 and the settings were first retrieved in February 1947.

It seems both machines were ‘lost’ in 1948 when the Soviets introduced emergency changes in their cryptologic systems.

It is not clear if Coleridge and Longfellow had a connection with the cipher machines that the Germans attacked during the war.

Coleridge may have been the system the Germans called ‘Bandwurm’.


Primary sources:

From TICOM I-2 ‘Interrogation of Dr. Huettenhain and Dr. Fricke at Flenshurg,21 May 1945’, p2

... they also have another machine„ funkfernschreiber, which encodes during transmission. it uses the international five impulse teleprinter code.

Q. Did the Russians use this machine much?

A. It became increasingly important during the last 1 1/2 years.

Q. What units used it?

A. Only the highest staffs, press and diplomatic services.

From TICOM I-30 ‘Report on Interrogation of  Uffz. Karrenberg  at Steeple Claydon on 7th July 1945 at 11.00 am’, p2

The subject of this interrogation was confined to Karrenberg’s work on Russian Baudot letter ‘strip’, traffic known to the Germans as ‘BANDWURM’ and not to be confused with Russian 5-letter, traffic also carried on Baudot lines.

The Germans had not captured any of the apparatus used but considered that it consisted of two parts: 1) a Baudot-teleprinter with the letters of the Russian alphabet (excluding q and 'g') and figure and letter shift making 32 characters in all; 2) a cipher attachment consisting of 5 small wheels driven by one large wheel.

Each of the small wheels had a pattern of positive and negative impulses and each wheel worked in conjunction with one of the five impulses produced by pressing a key of the teleprinter, the effect being to add a positive or negative impulse to each of the five impulses produced by the letter being sent. This in effect means adding a letter of key to the clear letter to produce a cipher letter.

Depths were frequent on the traffic intercepted by the Germans, but they do not seem to have made any attempt to reconstruct the wheel patterns. In the case of the driving wheel they came to the tentative conclusion that it had a period of 43. The preambles of messages were always enciphered which resulted in stereotyped and known beginnings to messages. The machine setting for a message was indicated by means of a two figure number which presumably referred to a table of settings; a different table was used each day.

Before the actual start of a message a passage of operator's chat was sent enciphered by the addition of a constant letter to each letter of the clear text. This letter was then sent en clair- and repeated three times. The object of this was to see that the receiver had his machine net up correctly.

The system was used by the Army and Air Force and to a lesser extent by the N.K.W.D.

From  TICOM I-153 ‘Second Interrogation of Uffz. Karrenberg of OKH on the Baudot-Scrambler Machine (Bandwurm)’, p2

P.W. stated that he believed the ‘Bandwurm’ traffic was first intercepted in 1940 in Warsaw. As far as he had been able to make out no interest had then been taken in it. The first actual knowledge we had of a traffic with the same external features (chat. indicators, eto.) was in summer 1943 when the first real interest was taken in it and the traffic was sent to Berlin for analysis. He understood that it went to a Dr. Pietsch and Doering.

There were a number of links usually varying according to the number of armies (Frontstaebe). The maximum number was 8. One end of each link was always in Moscow, the other would be mobile, and move with the armies. There were also one or two Airforce links. There was also supposed to be a link with the Far East. Traffic was heavier from the ‘outstations’ to Moscow.

and page 3

In Autumn 1944 both the end of 'adder' and every pause in the cipher proper was preceded by seven key letters [redacted]. Then the traffic went off the air and reappeared in December with no external change except that the seven ‘residue' letters had been reduced to three, suggesting a modification of the machine. In general it is clear that some of the features of the key at least have not changed in the last nine months.

From TICOM I-64 ‘Answers by Wm. Buggisch of OKH/Chi to Questions sent by TICOM’ , p2-3

1. Russian Systems 

In 1943 B heard that the Forschungsamt (no individual names given) had claimed some success on a Russian teletype machine, and had re-created the action of the machine. It was a machine with a very long cycle being not prime but the product of several smaller cycles--like the SZ42. B. did not know the cycle of all of the individual wheels or any other details. He heard this from DOERING, who was then doing his research on the T-52, but liaison with the FA was bad anyway (Major Mettig was particularly opposed to the SS taint) and the next he heard was that the traffic found by the FA had stopped. B. remembered only that the cycle of one wheel was 37; the others, he thought, varied widely from 30-80.

Late in 1943 and increasingly in 1944 OKH itself began to intercept non-Morse, 5-impulse traffic (called ‘Hughes’ by B.). The Mathematics Referat went to work on it, with TROEBLICHER playing a leading part. At the end of 1943 the Russians created a ‘kompromiss’, giving a depth of about 8 messages with the same setting. With this they were able to recover 1400 letters of pure key and at the same time to ascertain that the traffic being passed was the 5-figure code, with regular station chat enciphered at the same time on the machine (Suggests a machine in constant motion as described by Karrenberg).Part of the depth was created within the same long message, so that the machine had a cycle, at least in this one case of about 1450 letters. The actual number was thought to be very significant by the Germans, as it was prime and so could not be the product of smaller cycles in any way that they could imagine. This differentiated it from the machine which the FA had broken. The Germans postulated either a single tape machine like the T43 or a machine in which the motions of the wheels influenced each other,1 and 2 affecting 3, 3 affecting 5, etc. as in the T52. They were never able to prove one theory or the other. (B. apologized for this. Said they did not have enough mathematicians to tackle the fascinating problem of determining what the motion must be to create this cycle. Seemed quite convinced that there would be a unique solution to the problem.) After this experience they devised Hollerith machinery to locate depths, but in fact they only found three or four more cases and none of these gave additional cycle evidence or even furnished as much pure key as the first one. B. left the section in June 1944. He thinks the traffic slumped off in the summer of 1944 and LNA took steps to try to improve the reception, as they believed the traffic was still there. TROEBLICHER was detailed to this end of the work at this time. B. stressed one fact which had surprised him, that they had never had information about either of these machines (he assumed that the one the FA broke was not the same because of the difference on cycles.) from PW or agent sources.

B. said in passing that their own security idea on the subject of wheel machines of this sort was that the cycle should not be the product of smaller periods(as in Hagelin) even if this was long. Mutual influence of wheels should be used to avoid this, but at the same time care must be taken that too short a period was not created in the process. This in fact had apparently been done by the Russians, but the fact that it was not repeated suggested to him that they might have seen the weakness and corrected it.

Troeblicher is mentioned in CSDIC (U.K.) SIR 1717, appendix A as Troebliger (Uffz). He worked in Group IV, Referat 1b as an analyst on Russian Baudot traffic.


Sources: CSDIC (U.K.) SIR 1717, SI-32 ‘Special intelligence report’,  ‘European Axis Signal Intelligence in World War II’ vol2 and vol4, TICOM reports I-2, I-64, DF-98, I-30, I-169, I-153, I-173  , ‘The Secret Sentry: The Untold History of the National Security Agency’, Intelligence and National Security article: ‘Behind Venona: American signals intelligence in the early cold war’, Soviet cryptographic service 1920-1940   

TICOM reports I-2, I-64, DF-98 can be found in my TICOM folder, the other four can be downloaded from TICOM Archive.

Acknowledgments: It was Frode Weierud who first pointed to me the German interest in the Soviet teleprinter and the characteristics of the machine solved by the Forschungsamt in 1943. Credit also goes to Randy Rezabek for finding and uploading to the internet the interrogations of Karrenberg since they contain lots of information on the cipher machine.

No comments:

Post a Comment