Here’s a great challenge for my readers: Robert Bunch (1820-1881), a British spy in the US Civil War, left behind a number of encrypted messages that have never been deciphered. Can you solve this mystery?

Robert Bunch (1820-1881) was a British diplomat, who worked as a secret agent in Charleston, South Carolina, during the Civil War. Officially being a British representative following commercial matters, Bunch’s secret job was to inform his government about what was going on in the secessionist South of the US during the war. The British crown faced the dilemma that, on the one hand, they needed the cheap cotton they received from Southern State producers, while, on the other hand, they opposed the slave holding practices the US cotton production was built on.

The British even considered to mediate between the North and the South in the Civil War, but based on Bunch’s reports they realized that a solution acceptable for all parties was out of reach.

The story of Robert Bunch is told in the book Our Man in Charleston written by US author Christopher Dickey. An interview with Dickey about this work is available here.

 

Bunch’s encrypted notes

A few days ago, I received an email from Colin Beveridge, who had listened to my recent talk Solving Historical Ciphers with Modern Means at the History of Cryptography and Codes in London (organized by the British Society for the History of Mathematics). Colin has a very interesting website titled Flying Colours Maths. If you are looking for easy-to-unterstand information about mathematics, this is the place to go.

Colin wrote me because he had been contacted by afore-mentioned Christopher Dickey. When researching the life of Robert Bunch, Dickey encountered a number of encrypted notes Bunch had left behind. Apparently, nobody has been able to decipher it, so far.

Colin has provided me five cryptogram scans. Here they are (with transcriptions Colin made):

#1

Bunch-01

#2

Bunch-02

EOCH  TQOL  BTOU  QREH  DVTM  NRLF  DUTN  DRSK  MYHT  DMNT  QKYD  BYCV  PDUR  BOHK  RYEL  TLOT  BDVH  BQRY  RYBK  POFW  DELY  SEPT  LKOM  TMKN  XREH  HWTY  NLHZ  CPQV  BDUH

#3

Bunch-03

DUTM  CBFV  RQBH  MEOT  NKMQ  NDMO  SFWE  HLEK  RCUW  QREH  DMNT  RBUE  EKRS  QREH  CWRV  EQTY  SEPD  MNLE  ELLU  TMKN  WBEP  RISD  GMOS  RLEL

#4

Bunch-04

DUTM  UFVD  BHWK  RHEN  RDOY  LCVH  NRLD  QBOR  TMKN  TVDN  TMYH  TQSB  ECUH  LEML  RFLK  DMNH  YEOY  COMB  NLQS  TMKN  CDKM  NLOC  RQBH  RCNV  HPLE  DMNT  TWFL  RELY  QREH  SDLN  BQEW  EDLO  QREH  DUDN  NQOW  QREH  DCNT

#5

Bunch-05

 

Can you master the cypher?

The following letter is not encrypted but the author states that he has “mastered the cypher”.

Bunch-Mastered

Now, the question is whether a reader of this blog can master these encrypted messages, too. Encrypted messages from the US Civil War are not uncommon, as encryption played an important role in this war. The ciphers used back then were not the best, so virtually all known cryptograms from the Civil War have been deciphered. This one is probably more advanced than most other Civil War ciphers (I suppose that the British had better encryption methods than the US at that time), but it’s probably not unsolvable. My guess is that a polyalphabetical cipher (perhaps, a Vigenère variant) was used.

Can a reader find out more? If so, please let Christopher Dickey, Colin Beveridge and me know.


Further reading: Who can decipher this Pitman letter from the US Civil War?

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 (26)

  1. #1 Colin Beveridge
    WEYMOUTH
    2. Juni 2018

    I should mention that #1 *has* been cracked (that’s a substitution cipher – it says ‘stop the fools’, as is written at the top) – but is a significantly different cipher to the others.

  2. #2 Marc
    2. Juni 2018

    Looks like playfair

  3. #3 Thomas
    2. Juni 2018

    In Dickey´s book a similar cipher is shown which was decoded: https://books.google.de/books?id=2NUVBQAAQBAJ&pg=PT127. Unfortunately Dickey doesn´t mention his source and the type of cipher.

    • #4 Colin Beveridge
      2. Juni 2018

      #2 Marc: that was my first thought, too, but I’m almost certain it isn’t. The pair-counts for the first half of each block of four are completely different to those from the second halves. In fact, the letter counts for each position in the block are different, which seems to point to something like a Vigénère using a mixed alphabet.

      #3 Thomas: I can answer that! The previous cipher was the one used in picture 1; it’s a substitution cipher with several replacements for some letters. It was cracked because someone had considerately written the plaintext on the same sheet as the cipher 🙂

  4. #5 Max Baertl
    2. Juni 2018

    In #4 the group QREH appears 3 times

  5. #6 Michael Schroeder
    2. Juni 2018

    If you break the quad-grams up into bi-grams, you find no double digraphs (i.e., No AA or BB etc…). Further, the letter J is completely missing. This is all consistent with a Playfair cipher. Also, both the IC and digraph IC are consistent with Playfair. Playfair was invented around 1854 for use by the British diplomatic core.

    Everything points to Playfair. Any probable words?

  6. #7 George Lasry
    2. Juni 2018

    1) It’s not Playfair. My solver can handle very short messages,
    much shorter than those, and it failed on it.

    2) There are 5 repeated quadruplets (when looking at all the messagge together).

    DMNT 3
    DUTM 2
    QREH 6
    RQBH 2
    TMKN 4

    The rest are unique.

    I would guess that the simplest explanation, given the technology of the time, is a code
    book with 4-letter codes. Without access to the code book itself, it is impossible to solve
    such a small sample.

  7. #8 Michael Schroeder
    Berwick, Maine
    2. Juni 2018

    Here are the statistics for message # 4 from BION’s app:

    Winner is Playfair with 76 votes out of 100

    vote distribution:
    Playfair 76
    Two_square 9
    Four_square 7
    Bifid 3
    Bazeries 2
    Porta 1
    Seriated_pfair 1
    Patristocrat 1

  8. #9 George Lasry
    2. Juni 2018

    There is still the (low probability) possibility that this is a Playfair-like cipher (but not pure Playfair), as there are quite a few repeat bigrams. Here is the list, with two counts – # of appearances as the first bigram of a quadruplet, and # of appearances as the second.

    BD 2 0
    BH 1 2
    BQ 2 0
    DM 4 0
    DN 0 2
    DU 4 0
    EH 0 7
    EK 1 1
    EL 1 2
    KN 0 4
    LE 1 2
    LK 1 1
    LY 0 2
    NL 3 0
    NR 2 0
    NT 0 4
    OT 0 2
    OY 0 2
    PD 1 1
    QR 6 0
    RC 2 0
    RQ 2 0
    RY 2 1
    SD 1 1
    SE 2 0
    TM 5 3
    TQ 2 0
    TY 0 2
    UH 0 2
    VH 0 2

    It is interesting to see that a bigram usually appears either as the first or as the second pair
    but much less frequently as both.

  9. #10 George Lasry
    2. Juni 2018

    Just for the fun, I tested another Playfair-like hypothesis, that maybe the spy used Playfair deciphering instead of enciphering (and vice versa). No success.

    Any other suggestion for Playfair-like schemes?

  10. #11 George Lasry
    2. Juni 2018

    One more option is double Playfair.

  11. #12 Colin Beveridge
    2. Juni 2018

    I spent quite a lot of time looking at Playfair variants (including what George (#10) suggests, an alternating pair of grids) without any success. While there’s a lot of circumstantial evidence pointing to Playfair, I’m pretty sure it isn’t.

    I notice that, separating the counts for the 1st, 2nd, 3rd, and 4th letter in each block, each distribution is a) significantly different and b) consistent with a substitution.

    If it turned out to be a codebook, I’d be very disappointed! :o)

  12. #13 Narga
    2. Juni 2018

    I don’t always agree with the transcription. Notably there seems to be another letter (looks like J) that seems to be mixed with T here. It is quite common, so it might impact the analysis. I’m talking about the first letter of the second group in #2. It could be a new letter “J” or I or S, but it is so consistent that I’d call it J.

  13. #14 Hias
    2. Juni 2018

    Mir ist aufgefallen, dass zwei verschiedene Symbole als „T“ bezeichnet werden. Eines ist aber wohl eher als „J“ zu lesen.
    Z.B. bei Text #2 Wort Nr16: Das müsste JLOT nicht TLOT heißen. Oder Wort Nr2 JQOL anstatt TQOL.

  14. #15 Colin Beveridge
    2. Juni 2018

    Thanks, Narga and Hias — I am quite willing to believe my transcription is incorrect in places! J looks plausible to me.

  15. #16 Hias
    2. Juni 2018

    @Narga #13
    Sorry, I see you notice the letter „J“ faster than me.

  16. #17 Narga
    2. Juni 2018

    @Hias: I think we posted it at about the same time.
    @Klaus: Thanks for the quick edit.

    Does anyone have an idea what the second letter in the second line of #5 is?

  17. #18 Colin Beveridge
    2. Juni 2018

    @Narga #17, looking at this: https://www.pinterest.co.uk/pin/469781804855200091/ , it most closely resembles a C (but I’m not entirely convinced).

  18. #19 Marc
    2. Juni 2018

    Some characters looks like digits, perhaps playfair 6×6 ?

  19. #20 Klaus Schmeh
    2. Juni 2018

    Blog reader Matthias via email:
    Bei Bild 2 ist mr aufgefallen, dass noch ein Text von der Rückseite durchscheint.
    Habe Ihnen ein überarbeitetes Bild angehängt, bei dem der Text einigermaßen zu sehen ist.

  20. #21 Colin Beveridge
    2. Juni 2018

    I’m afraid it’s not trivial to get a picture of the reverse – however, I did my best to flip and pick out the text to transcribe it. I have:

    BDUH NLQT MDTU DLQH YKED WBEP ALST SQYT RAFW

  21. #22 Norbert
    3. Juni 2018

    The text on the back of letter #2 gives us two more repeated four-letter-groups: BDUH and WBEP. Of these, BDUH can be ignored, because its counterpart is written in the bottom right corner of the front side of the same letter, obviously to indicate the continuation, as in old printings.

    However, the frequency of repeated four-letter-groups is striking. Much less frequent are repeating quadgrams overlapping two groups: “.REH D…” appears four times, “…Y SEP.” and “…W QRE.” twice, and I found only one repeated quadgram at different positions in the four-letter-pattern, “MYHT” in #2 and “.MYH T…” in #4.

    All in all, I would agree with George on the code book hypothesis.

    (If not a code, it could imho only be a periodical polyalphabetical substitution with four indepent alphabets, but my hill climber, based on this assumption, has not yielded any meaningful result so far.)

    • #23 Colin Beveridge
      3. Juni 2018

      Thanks, Norbert (#22) — mixed ciphers with a period of 4 is the conclusion I came to after a lot of messing about with Playfairs and so on (obviously, I didn’t get anywhere, either! – but the count statistics look about right) I wondered about some rearrangement being put on top of a cipher, but figured that was probably more sophisticated than I could expect from these clowns ;o)

  22. #24 Norbert
    4. Juni 2018

    Although the letter frequencies seem to indicate a mixed cipher with four alphabets, two arguments speak against it: First, we should expect repeated quadgrams to distribute evenly over the possible block positions. But there are only 3 repetitions (i.e., 4 occurences) of REHD at positions 2341, two repetitions at 4123, and none at 3412, as opposed to 12 or 13 repetitions at 1234.
    Second, with a mixed cipher, there is no need to end every message with a full four-letter-block (and, moreover, there is no evidence that the last block was filled up with a special letter).

  23. #25 Narga
    4. Juni 2018

    @Norbert (#24): In addition, with a four alphabet cipher you would expect a lot more double letters (KK, BB, …) in the enciphered text. There are almost none of those. I find the general absence and the few presences (…E ELLU) one of the strongest indications for a bigram-substitution very close to Playfair.

    Didn’t you have a more general SimAnneal code for bigram substitutions that broke the challenges by Klaus? https://scienceblogs.de/klausis-krypto-kolumne/2017/02/13/bigram-substitution-an-old-and-simple-encryption-algorithm-that-is-hard-to-break/
    But probably this one here is not long enough…

  24. #26 Colin Beveridge
    12. Juni 2018

    I’ve put together a summary of arguments for and against each of the hypotheses I’ve heard mentioned here — happy to update and/or correct as needed.

    Playfair

    For:
    Lack of double letters
    Even-length messages in blocks of 4
    Repeated strings generally start at column 1

    Against:
    No solution found
    Statistics for pairs at column 1 wildly different from column 3
    Statistics for first halves of pairs different to second

    Playfair variant

    For:
    As for Playfair
    Against:
    Doesn’t account for different statistics across pairs
    Too complicated for these clowns?

    Codebook

    For:
    Repeated “words” in same place
    Nothing else seems to work
    Against:
    Statistics of each column look alphabetical
    Bunch explicitly says ‘cypher’

    Vigenere

    For:
    Certainly seems periodic
    Against:
    No solution found

    Periodic multiaphabet

    For:
    Fits with column-by-column statistics
    Against:
    No double letters
    No solution found
    Repeated quadgrams not evenly distributed