Im Jahr 1920 ließ der deutsche Textilingenieur Rudolf Zschweigert eine Verschlüsselungsmaschine patentieren. Heute stelle ich einen Text vor, der mit dieser verschlüsselt wurde. Kann ein Leser sie lösen?

English version (translated with DeepL)

Letztes Jahr habe ich für die HistoCrypt einen Beitrag über die Verschlüsselungsmaschine von Rudolf Zschweigert eingereicht. Diese Arbeit wurde zwar angenommen und ist dann im Konferenzband erschienen, doch die Veranstaltung selbst ist Corona-bedingt ausgefallen. Den geplanten Vortrag über das besagte Gerät konnte ich daher nicht halten.

Am kommenden Samstag (20. März 2021) habe ich die Gelegenheit, die abgesagte Präsentation nachzuholen, und zwar als Webinar im Rahmen der International Conference on Cryptologic History (ICCH). Es handelt sich bereits um meinem siebten Vortrag in dieser Reihe. Er findet dieses Mal um 17 Uhr (also eine Stunde früher als sonst) statt und ist (wie immer) kostenlos. Die Zugangsdaten werden über die Mailing-Liste verschickt, die man auf der ICCH-Webseite abonnieren kann. Wer dieses Abo nicht will, kann mir gerne eine Mail schicken, dann lasse ich ihm/ihr den Link zukommen.

Mein Partner bei diesem Vortrag ist der Brite Jerry McCarthy, über den ich auf Cipherbrain bereits mehrfach berichtet habe. Jerry hat unter anderem die Enigma-Knackmaschine “Zyklometer” nachgebaut.

Quelle/Source: McCarthy

 

Der Zschweigert-Kryptograf

Die von Zschweigert entwickelte Verschlüsselungsmaschine (Zschweigert-Kryptograf) ist einer der vielen Chiffriermaschinenentwürfe, die in den Jahren nach dem Ersten Weltkrieg (1914-1918) entstanden sind. Zschweigert war ein Textilingenieur, der zuvor Näh- und Webmaschinen konstruiert hatte, wovon weitere Patente zeugen, und dabei offenbar auf die Idee kam, ein Chiffriergerät zu bauen.

Quelle/Source: Schmeh

Anders als die Enigma und so gut wie alle anderen mechanischen Kryptogeräte realisiert der Zschweigert-Kryptograph eine Transpositions-Chiffre, und nicht etwa eine Substitutions-Chiffre. Buchstaben werden hier also nicht ersetzt, sondern in der Reihenfolge vertauscht.

Die Zschweigert-Maschine arbeitete mit neun parallelen Papierrollen. Der Schlüssel wurde auf einer neunspaltigen Lochkarte bereitgestellt. Die Zahl der Lochreihen war beliebig. Pro Reihe gab es genau ein Loch. Die Maschine arbeitete den Klartext Buchstabe für Buchstabe durch und sprang dabei jeweils eine Reihe auf der Lochkarte weiter.  Die Spalte, in der sich das Loch befand legte fest, auf welche Rolle gedruckt wurde. War die Maschine in der letzten Reihe angekommen, dann sprang sie anschließend wieder an den Anfang.

Das folgende Video zeigt, wie die Maschine funktionierte (die Lochkarte hat hier fünf Reihen):

Wer es genauer wissen will, kann meine HistoCrypt-Arbeit oder das Patent aus dem Jahr 1920 durchlesen. Dieses Patent ist die einzige mir bekannte Quelle zum Zschweigert-Kryptografen. Im Internet habe ich zwar weitere Informationen zu Rudolf Zschweigert und dessen Unternehmen gefunden, doch die Maschine wird nirgends erwähnt. Es ist zwar möglich, dass sie gebaut wurde (als Prototyp), ich gehe aber davon aus, dass sie nie in der Praxis eingesetzt wurde.

 

Jerrys Zschweigert-Simulation

Der besagte Jerry McCarthy hat auf Basis meiner HistoCrypt-Arbeit eine Computer-Simulation des Zschweigert-Kryptografen programmiert. Hier ist ein Screenshot:

Quelle/Source: McCarthy

Im Vortrag am Samstag wird Jerry diese Software vorstellen.

 

Eine Challenge

Wie sicher der Zschweigert-Kryptograf ist, hängt vor allem von der Schlüssellänge ab. Diese wird von der Anzahl der Reihen der Lochkarte festgelegt. Im Video hat diese Karte nur fünf Reihen, was recht knapp bemessen ist. Es sollte jedoch möglich sein, auch deutlich längere Lochkarten einzusetzen.

Um etwas mehr über die Sicherheit dieses Geräts herauszufinden, habe ich eine Challenge erstellt. Ich habe zu diesem Zweck einen englischen Text mit knapp über 100 Buchstaben mit dem Verfahren des Zschweigert-Kryptografen verschlüsselt. Das Ergebnis sieht wie folgt aus:

Quelle/Source: Schmeh

Spaltenweise ausgelesen erhält man folgenden Geheimtext:

EAEA-BTABOASIIE-MOCNSTSMLLNIWSNNICE-RYTKSNHTERYTTWKHPPH-YHWTYIFRR-NTIOOKACNHCGIATR-DLTRDOTS-EOOREPEDNCAN-CPIECPRI

Die Anzahl der Reihen auf der Lochkarte ist Teil des Schlüssels.

Kann ein Leser dieses Kryptogramm knacken? Falls es jemandem bis zum kommenden Samstag gelingt, werde ich dies natürlich im Vortrag erwähnen.


Further reading: Prince Edward opens new exhibition at Bletchley Park

Linkedin: https://www.linkedin.com/groups/13501820
Facebook: https://www.facebook.com/groups/763282653806483/

Subscribe to Blog via Email

Enter your email address to subscribe to this blog and receive notifications of new posts by email.

Kommentare (10)

  1. #1 George Lasry
    Givataim
    19. März 2021

    The plaintext is most certainly

    MYENCRYPTEDBOOKSLISTCONTAINSBOTHWORKSTHAT
    ARECOMPLETELYENCRYPTEDANDWORKSINWHICHSIGN
    IFICANTPARTSAREINCIPHER

    and the key length is probably 51, but I am having some trouble finding an exact key that reproduces the plaintext accurately.

  2. #2 Narga
    19. März 2021

    Great work again, George! My code somehow found less words than I could see with my bare eyes 🙂

    Very interesting challenge! Already from the first glance one suspects “CIPHER” to be the last word and from the length of the columns one gets either 24 or 51 as the most likely key length. With 51 automatically “ER” at the end and therefore “MY” at the beginning are fixed as these are the columns with an odd number of letters and you can apply the full key twice. It feels like you can do more with pen and paper here than with software.

  3. #3 Norbert
    19. März 2021

    Great job, George! I suspect that the plaintext reads “MYENCRYPTEDBOOKLIST…” (one “S” dropped of your solution),
    and the 51 first key elements could be

    3,5,1,6,9,4,4,9,4,8,
    7,2,3,8,4,7,9,4,2,3,
    8,3,6,2,6,4,3,2,6,3,
    5,5,6,8,6,3,5,4,1,7,
    6,7,8,6,2,3,8,3,1,4,
    4

    but it does not really work as key length 51. After 65 letters I find no possibility at all to read the plaintext out, so maybe it is still a bit incorrect, or could Klaus’ plaintext have some typos?

  4. #4 Klaus Schmeh
    19. März 2021

    @George: Congratulations! Your solution is correct.
    As we already discussed by mail, there is a typo in the transcribed ciphertext (the ciphertext in the diagram appears to be correct): The second T in the TT in the fourth word should be deleted. Sorry for this. I’m glad, you solved the challenge anyway.

  5. #5 Klaus Schmeh
    19. März 2021

    Here’s the key:

      123456789
    01--X------
    02----X----
    03X--------
    04-----X---
    05--------X
    06---X-----
    07---X-----
    08--------X
    09---X-----
    10-------X-
    11------X--
    12-X-------
    13--X------
    14-------X-
    15---X-----
    16------X--
    17--------X
    18---X-----
    19-X-------
    20--X------
    21-------X-
    22--X------
    23-----X---
    24-X-------
    25-----X---
    26---X-----
    27--X------
    28-X-------
    29-----X---
    30--X------
    31----X----
    32----X----
    33-----X---
    34-------X-
    35-----X---
    36--X------
    37------X--
    38---X-----
    39X--------
    40----X----
    41-----X---
    42------X--
    43-------X-
    44-----X---
    45-X-------
    46--X------
    47-------X-
    48--X------
    49--------X
    50---X-----
    51---X-----
  6. #6 George Lasry
    19. März 2021

    Thanks, I wrote a program that can find solutions in minutes with simulations, but I only got fragments with length = 51. I was finally able to collate into a coherent full anagram, that constituted the plaintext, but could not find a matching key. With the help of Klaus, we found some error in the ciphertext, and everything makes full sense now.

    Anyway, that error made the challenge even more interesting …

  7. #7 Gerry
    19. März 2021

    The numeric key is
    3 5 8 6 9 4 4 9 4 1 7 2 3 8 4 7 9 4 2 3 8 3 6 2 6 4 3 2 6 3 5 5 6 8 6 3 7 4 1 5 6 7 8 6 2 3 8 3 9 4 4

  8. #8 Klaus Schmeh
    19. März 2021

    @Norbert, Gerry, George, Narga:
    Thank you very much for your contributions and sorry for the typo. The challenge is solved now. I hope to see some of you in the talk tomorrow.

  9. #9 Klaus Schmeh
    29. März 2021

    Bill Ricker via Facebook:
    Sorry i missed real-time watch & comment; catching up now.
    Jacquard loom connection — pattern weaving is not unlike transposition!
    Scarcity of (pure) transposition machines – I’m glad you said “until 1970s,” since most modern computerized algorithms (at a minimum, any Feistel and similar Block ciphers) blend substitution (confusion) and transposition (diffusion).
    At least one cusp-of-modern WW2 German systems was also blended: S&H T52 Geheimschreiber SFM, known to Bletcheley as STURGEON (of the FISH family of teleprinter ciphers) had both Vernam-style substitution (bitwise XOR-ing) and transposition (bits swapped among the 5 parallel bit-streams). It was _inadequate_ diffusion in that the diffusion only changed the effective substitution of one character; the block size was still only 5 bits, with or without bit swapping.
    Hypothetically, there was one letter handling machine that had storage: a purely mechanical Linotype hot-metal typesetting machines had a circulation and storage system for the matrices. I would expect there’s a patent somewhere for abusing those mechanisms to doing keywise anagram of each line before casting a line and then distributing the matrices for reuse. If one automagically cast line numbers from a key into each slug, that could serve as an instruction to manually reorder the slugs before taking a proof-print (or reading the type in a mirror!) (That’s still single permutation, that’s just reading out of the matrix by a keyed order of rows.)
    (Would need to use a monospace font-face to make this work sensibly.)
    To do proper double permutation (with two different sized tableaus) one would need to rekey the output of the first permuting-Linotype on a second one with different width and keying.

  10. #10 Klaus Schmeh
    30. März 2021

    Bill Ricker via Facebook:
    I don’t know that anyone ever had the thought before. Might be worth trolling the Patent Office indexes? (Alas a simple search gets encryption as a copy protection for commercial Linotype font designs. I don’t quickly find any Linotype encryption patents before 1980.)
    Your comment on mechanisms with storage reminded me of where i’d seen an automated mechanism with temporary working storage for a line of letters, automated recycling of them, and a very simple adjustment (e.g. adjusting word-spaces for flush margins), which convinced me that a decent transposition machine was possible with late Victorian / Edwardian technology, and (sh|c)ould have been obvious to someone before me.
    Since it doesn’t seem to have existed, maybe i’ll give the idea to the authors I know working in Alternate Histories .
    Alas as we know from WW1 histories, even double transposition is subject to cryptanalysis if there’s enough traffic “depth” under a single key … and mechanizing implies enough traffic, so it’s not really a great solution.
    (The Linotype i suspect HAS been used to do so-called ^Baconian^ stegano-encipherment in the examples in proponent books, since shifting between regular & bold (_or_ regular & italic, depending on the font loaded) is a shift. Would still be charged as penalty copy like mathematics or chemistry … but would hurt less on a Linotype than on a Monotype or hand composition, which latter is what the proponents alleged Bacon had paid the printers for. Having done hand comp, that’s risable. Nope, the printers of the First Folio had sloppy apprentices who’d distributed non-identical “sorts” into the case.)