Unknown 4-23-2019

KWVYI OKSGQ MCETS LECCK NNLBD GOSAX EWGRN RMDXO SBGAW AKGBN SFUBE OLMSR HGGEL MHHRI OBXDO RAIDS HKIUE MXIEX IPEWC HUNPA YARCH

Hint: SHULRGQLQH ends: DQGFULHGDK

The pt is from two unrelated Gutenberg texts. My hillclimber solved this without the crib, but it took over 500 million trials. I had to let it run overnight. My analyzer identified the type correctly by a good margin.

14 thoughts on “Unknown 4-23-2019”

  1. The second passage continues: MJBNQQSTYKTWLJY

    My random forest test put the correct type in second place. Both of my neural net tests put the correct type on top. I tried PH hill-climbing without a crib for awhile but didn’t solve it. Usually my javascript hill-climber solves this type quickly without a crib — I haven’t even put a crib routine into it. I decided to use the crib and my old wimpy hill-climber written in C solved it almost instantly using the crib. I went back to my javascript PH hill-climber and tried to solve it again without the crib. The PH hill-climber finally solved it without the crib in about 10 million trial decrypts while using an unusually low fudge factor — the fudge factor was so low that the hill-climber was accepting less than one percent of the keys it generated.

  2. Any chance you guys can run your analyzers on the (possible) cipher discussed here:

    https://www.reddit.com/r/codes/comments/beq53k/mystery_text_found_written_in_a_library_book_is/

    Transcript:

    YMBFPTYSYSNAOSBYAACPTM
    YSNEMTMOIYJGTEBYNIAJ
    APSNDETMYLGAIdISYMI
    HFBYTFAAWOTYCMCHNEFTY
    IHSAWOYYHNEAWOWIWLTDI
    HAJBSASTFYTMMFdOETTD
    ITWYGMTIPDTTOFYANRSITL
    OACYANWYAJAHCTDSIHND
    ONIPEICNCOACOHMOOOT
    YPMDISWtOKIEYWOP,R,A
    AOTWFFFYOTYOMIHLYD
    WICSTOTIhhNCFBYHNLAO
    WALYCWOWTDSOLIOMWWW
    YNTLTYMYFYLNAWNMWSSI
    NDIDCWGOOBOYAYOFW
    GYKODLIAALACNSuFMSIAt
    OJBaHTYIANaWJSWDET
    OMYHABEODIASMGWIhdt
    TOTYHNdTIAPASYOAYdTS
    WBAAF,YFNSmTMOTOHYA
    NSYJDWMABTdEKMMSA
    IHO,M,LTPYANWADNWTSWIS

    AIHNGTSTBFANTdWL.

    IAAYYAARYNTTMIFLYTIA
    BYWYEWGOTAYNTTMWITA
    COSSOCAYSLIYATMTTOTT
    NAAMDOTWYNSOETBRNBAH
    TMAYSIDKINADRHBOFTW
    TTFBWSYJSBYKIWBMATO
    GTDALAYANIATADWCBYD
    DLOAHWIAFHWSTAYSTF
    OHSELTODTNOAMaHMA
    HWCUALFMTFYM.YNWTGTO
    G.AHTCHASOD.ATRISWNB.
    WISIAAOHB,BYCSTSIPYAA
    FITIYWSTASIABBaagi
    WSYCIATOYSIAWyd,BHS
    AIIMYMMFUIOIYRYCH
    CLEFAOWNHOTSADIHIW
    YCTTACBIWLFYTTiiM,L
    PIDOE,GSSIiMaTCIDLS
    DMSFHATHTLTIFTNWIA
    TOYLMJOMATLOIMFJMW
    YTOOMdYLDBLPCEAT

    One theory, based on the letter distribution, is it’s a mnemonic (each letter is the first letter of some word).

    There are four letters missing from the cipher text: Q, V, X and Z.

    I tried out some of BION’s online ID tests. Neural net top three types were:

    RouteTransp with 11 votes
    Redefence with 7 votes
    Myszkowski with 3 votes

    Random forest test top three:

    Redefence 21
    RouteTransp 15
    Patristocrat 13

    What do you think?

    1. I don’t think it’s a transposition type. The Normor score is too high. The W and Y are too frequent. The index of coincidence is consistent with normal English and a Patristocrat came out on top in my Analyzer with a long stretch in the middle of this. It’s not clear what to do with the lower case letters. My program isn’t designed to analyze anything like that since no ACA cipher mixes them. There’s no periodicity so that leaves out a lot of types. If I had to guess I’d go for either a hoax or a Null.

  3. I tried a neural net collection that uses all of my experimental stats and it got:
    Top cipher types were:

    Amsco with 14 votes
    Patristocrat with 14 votes
    CheckerBoard with 2 votes

    Total votes: 30

    But, as RAT says, it has too many Y’s (74) and W’s (52) and too few E’s (17), to be convincing as a transposition cipher. And it doesn’t have enough repeated digraphs to be convincing as a simple substitution cipher.

    The mnemonic theory seems as good as any.

  4. Thanks for taking the time to analyze it! Guess it remains a mystery for now, unless something pops out if someone tries any null/concealment patterns. Or if there’s a convincing source for the mnemonic.

    Here’s my letter frequency stats comparing the distributions:

    https://docs.google.com/spreadsheets/d/1Tu2CnIPUNj6tlMlToFVCFcDfivEmJ7qMAkaV4VGVLAQ/edit?usp=sharing

    Ys, Ws, and Es do seem somewhat in line with the expected distribution for a first-letter mnemonic. I have a 500GB compressed corpus of text for these sorts of tasks so I wrote something that samples them for sequences that might match. Nothing really interesting popped out, but it didn’t scan the entire corpus (it would take forever). Best matches I got were things like:

    ALAYANIATAD: AS LONG AS YOU ARE NOT IN A TRIPLE-A DESKTOP
    HWIAFHWSTA: HE WAS INCHES AWAY FROM HER WHEN SHE TURNED AROUND
    IYATMTTOTT: IF YOU ARE TALKING MMO THEN THE ONLY THINGS THAT

    which are probably just happening by chance.

    For all we know, the source of the mnemonic might the very book it is written in.

  5. I had fun writing a program to find acronym sequences. I have a batch of a few dozen books from gutenberg and tried searching the beginning of this ciphertext (YMBFPTYSY). The longest stretch my program found was the first four four letters:
    ymbf what do you mean by fun
    ymbf new york must be filled
    ymbf you must blind fold me
    ymbf young man bounce from a
    ymbf he’s ferce you must be fercer
    ymbf you must be for there is
    etc.
    I don’t buy that this text is is mnemonic for a series of words. The frequent Y’s are not consistent with most texts I’ve found, although if the words YOU, YOUR are frequent in the text, I suppose it’s possible. Checking both this Gutenberg corpus and my solutions to ACA ciphers, the Y places 18th in frequency as an initial letter.

    1. It’s a fun exercise. Mine found:

      YOU MUST BE FROM PASADENA TAKING YOUR
      YOU MAY BENEFIT FROM POSTING THE
      YOU MEAN BY FLAVOR PROFILE THOUGH
      YOU MUST BE F’ING POOR THEN YOU

      The Y is definitely an anomaly. Maybe the story it comes from has several characters whose names starts with Y. 🙂

  6. One more observation on the cipher text: There are many sequences that match with an edit distance of one.

    Examples:

    ARYNTTM AYNTTM
    YNTTMI YNTTMWI
    DETMY DETOMY
    TOFYAN TOHYAN
    YANWYA YANWA
    ARYNTT AYNTT
    RYNTTM AYNTTM
    RYNTTM YNTTM
    YNTTM AYNTTM
    YNTTM YNTTMW
    YNTTMI YNTTM
    YNTTMI YNTTMW
    NTTMI NTTMWI
    WSTAYS WSTAS
    CIATOY WIATOY
    CIATOY IATOY
    IATOY WIATOY
    IATOY IATOYL
    IATOYS IATOY
    IATOYS IATOYL
    PTYS PTMYS
    TMYS TMAYS
    YNIA YANIA
    NIAJA NIATA
    SNDET SWDET

    Each pair occurs in the cipher text, and each element of the pair only differs by one edit operation.
    But, it is unclear to be if these are any significance. There are about 500 such pairs in the cipher text, and even after shuffling the cipher text you can find about 400 similar pairs. But maybe the longer sequences are more significant.

    Here are some longer sequences that differ by two edit operations:

    YHNEAWOW YHNLAOW
    TOTTNAAM TODTNOAM
    YNTTMIF YNTTMWIT
    YNTTMI YNTTMWIT
    YNTTMI AYNTTMWI
    RYNTTMI AYNTTMWI
    ARYNTTMI AYNTTMWI
    ARYNTTMI AYNTTMW
    ARYNTTMI AYNTTM
    AARYNTTM AYNTTM
    AARYNTTM TAYNTTM
    SNDETMY SWDETOMY
    TOFYANRS TOHYANS
    HNEAWOW HNLAOW
    YHNEAWO YHNLAO
    YHNEAW YHNLAOW

    Could these be an indication of some kind of structure or meaning, or do they just arise randomly?

  7. I experimented with a simple hillclimber that removes randomly selected letters from the cipher text in order to maximize the production of valid word sequences, under the premise that a real message is hidden among null cipher text letters.

    Here’s a sample of what it found towards the beginning of the cipher:

    https://pastebin.com/raw/pCxzCqC0

    Probably just nonsense, and concealments would probably follow a less random pattern, but it was fun to see the climber trying to make words.

Leave a Reply

Your email address will not be published. Required fields are marked *