I’ve just had a particularly interesting email exchange with Paul Relkin concerning the Feynman Challenge Ciphers, which he has generously allowed me to share here. The context is that the first Feynman Challenge cipher’s plaintext was from the very start of Geoffrey Chaucer’s Canterbury Tales, i.e. the first twelve lines of the General Prologue:

WHAN THAT APRILLE WITH HIS SHOURES SOOTE
THE DROGHTE OF MARCH HATH PERCED TO THE ROOTE
AND BATHED EVERY VEYNE IN SWICH LICOUR
OF WHICH VERTU ENGENDRED IS THE FLOUR
WHAN ZEPHIRUS EEK WITH HIS SWEETE BREFTH
INSPIRED HATH IN EVERY HOLT AND HEETH
THE TENDRE CROPPES AND THE YONGE SONNE
HATH IN THE RAM HIS HALVE COURS Y-RONNE
AND SMALE FOWELES MAKEN MELODYE
THAT SLEPEN AL THE NYGHT WITH OPEN YE
SO PRIKETH HEM NATURE IN HIR CORAGES
THANNE LONGEN FOLK TO GO ON ON PILGRIM[AGES]

Paul writes:


The Prologue

I’d like to share with you a possible clue I’ve discovered to the sources of the 2nd and 3rd Feynman Ciphers. My findings relate to the identification of a specific published transcription of the Canterbury Tales that is the probable source of the 1st Feynman Cipher.

As you are probably aware, the Canterbury Tales have been transcribed and reprinted innumerable times. Among the many different published editions of the Canterbury Tales, there are several idiosyncratic spellings associated with particular transcriptions. Although individual lines are spelled the same way in many different editions, I found that the 12 lines of the Feynman Cipher taken together are unique enough to match only one published transcription, like a “word fingerprint”.

To find the edition that the Feynman Cipher is based on, I extensively searched for editions of the General Prologue that were published before or during World War II and compared the word spellings to the Feynman Cipher.

First, I discovered what may be a typo in the 1st Feynman Cipher. The word “brefth” does not appear in any published edition of the General Prologue I have been able to identify. The most likely correct spelling is “breeth”.

Second, I found that the only version of the General Prologue that matches the Feynman Cipher is Fred Norris Robinson’s 1st edition of Chaucer’s Complete Works. In the introduction to his book, Robinson actually discusses several of the uniquely spelled words that later found their way into the 1st Feynman Cipher and explains why he rejected the popular spellings and chose less common ones.

Possible Sources

Having identified Robinson’s transcription as the probable source of the 1st Feynman Cipher, I discovered that there are only a few different editions of this transcription that were published between 1933 and 1938 that could have been used by the author of the Feynman Ciphers:

In 1933, Houghton Mifflin published this book in at least three editions:

The Complete Works of Geoffrey Chaucer (black):

The Complete Works of Geoffrey Chaucer, Student’s Cambridge Edition (red):

The Poetical Works of Chaucer, Cambridge Edition (white):

In 1936, Houghton Mifflin published small books containing parts of Robinson’s Canterbury Tales with an introduction written by Max John Herzberg. The title of the book that contains the quote used in the cipher is “The Prologue, the Knight’s Tale, and the Nun’s Priest’s Tale”:

In 1938, Houghton Mifflin included Robinson’s Canterbury Tales in a two volume collection of British poetry by Paul Robert Lieder called “British Poetry and Prose” (Volume 1):

Interestingly, Robinson’s 2nd edition of Chaucer’s Complete Works in 1957 no longer matches the spellings in the cipher!

It’s specifically here where I think we may find clues to the 2nd and 3rd ciphers. It seems plausible to me that “British Poetry and Prose” contains other literary works that were the basis for the 2nd and 3rd Feynman Ciphers. For example, several of its poems have 6 letter words that repeat twice, consistent with “CJUMVRCJUMVR” in the 2nd Feynman Cipher.

Robinson’s 1933 book of Chaucer’s Complete Works could also be the source of the 2nd and 3rd ciphers. The 1933 book is part of a series of books called “The Cambridge Poets” and the 1936 book is part of a series called “The Riverside Literature Series”. The other books in the series are also potentially worth looking at.

Los Alamos?

My research suggests that several copies of these books have the original owner’s name and other notes written in them. If we were able to locate the copy that was used at Los Alamos, it might reveal the name of the scientist who created the ciphers. There may be other writings within it that would give further clues about the ciphers.

I discovered that the Mesa Public Library in Los Alamos has a copy of Robinson’s 1933 book. The Mesa Public Library originated during World War II in the Big House where Feynman lived, so I wondered whether the library book could be the copy that was used to create the cipher.

So, I recently arranged to borrow that book through interlibrary loan. Since I live on the East Coast, I had to try 5 different libraries before I found one that would let me request that particular book. It then took two tries because they accidentally requested the book from the Mesa Public Library in Arizona instead of the one in New Mexico. I finally received the book I requested. Unfortunately, the book plate indicates that it was donated to the library in the 1970s. This makes it unlikely (albeit not impossible) that this was the specific copy used in the period around World War II to create the 1st Feynman Cipher.

I hope you find this information interesting and that it brings us a step closer to solving the 2nd and 3rd Feynman Ciphers.

Chaucer and Cryptography?

(((NickP: I responded here, pointing out:)))

Incidentally, there are two interesting links between Geoffrey Chaucer and cryptography. The first (which you may well have heard of) is that he included six blocks of ciphertext in his Treatise on the “Equatorie” (basically a kind of astrolabe). But the second is that a very major work on Chaucer (finally published in 1940) was written by John Matthews Manly and Edith Rickert, both well-known code-breakers. (I’ve covered them a few times on CM, mainly because of Manly’s links to the Voynich Manuscript.)

However, Rickert died in 1938, Manly died in 1940 and Los Alamos only really started in 1943, so we can rule out a direct transmission from either of them to Feynman. All the same, I do consider it entirely possible that one/both of them was/were the ultimate source of the three cryptograms. Just so you know!

(((To which Paul replied:)))

Concerning your excellent point about Rickert and Manly, there was another colorful link between a Chaucer scholar and Los Alamos that I found while I was researching editions of the Canterbury Tales. John Strong Perry Tatlock was a famous Chaucer expert who transcribed Chaucer’s Complete Works. His daughter, Jean Frances Tatlock, had a romantic relationship with J. Robert Oppenheimer between 1936 and 1939. They continued to have an affair during Oppenheimer’s marriage. Their relationship was used as evidence against Oppenheimer during his security clearance hearings because Tatlock was a member of the Communist Party. As you know, Oppenheimer and Feynman had more than a passing acquaintance – as for Tatlock and Feynman, who knows?

Jim Lyons has returned to battle against the unsolved Feynman Ciphers: but this time round he’s wondering whether one or more might employ some variant of the Hill cipher.

It’s possible but… given the fact that #1 was a straightforward transposition of Chaucerian English, I don’t honestly buy into the idea that the others will prove to be cryptographically exotic.

To my mind, whoever set the first cipher seems (if the much-repeated back story itself is not itself a jest) to have been far more interested in snickering into his beard about having pulled the wool over Richard Feynman’s sainted eyes than proving his depth of cryptographic reading. I’d agree he could conceivably have wheeled out a Hill + substitution cipher crypto mechanism, but surely the meta-point of the whole exercise was that it was supposed to be a Los Alamos in-joke at Feynman’s expense?

Los Alamos

The Feynman Ciphers surfaced on Usenet in 1987 while Feynman was still alive (though he died in 1988), so it seems fairly unlikely to me that these were composed then. Hence it seems likely to me, on the balance of probability, that they did come from his time at Los Alamos: perhaps someone who was there with Feynman might remember?

There’s a nice page full of Feynman’s reminiscences of his time there 1943-1945, but that didn’t immediately answer the question.

All the same, this quickly led mw to the very watchable Memoir of Los Alamos in World War II with Murray Peshkin on YouTube. Given that Peshkin worked with Feynman and is still very much alive, I thought it worth a shot asking if he remembered the appearance of any ciphers. So I emailed him. 🙂 His response:

This is the first I hear of the Feynman ciphers. Of course I looked the question up, but nothing I saw related to anything of which I know.

Sorry not to be helpful

Oh well… if you don’t ask, you don’t find out.

The British Mission

However, given that the plaintext to the first Feynman Cipher was from Chaucer’s Canterbury Tales, it also struck me that the encipherer might well have been British. There was a sizeable British Mission at Los Alamos: the British had been working on an atomic research programme codenamed ‘Tube Alloys’ for some time, so had a bit of a head-start in the whole blowing-up-the-world race thing.

I couldn’t find a reasonable list of the British Mission personnel online, so decided to put one together: and here it is. If you have better biographies or links for any of the unlinked scientists, please let me know and I’ll update them here.

The British Mission to Los Alamos:
* James Chadwick (head of the mission)
* Egon Bretscher
* Boris Davison
* Anthony P. French
* Otto Robert Frisch
* Klaus Fuchs
* James Hughes
* Derrik J. Littler
* William G. Marley
* Donald G. Marshall
* Philip Burton Moon
* Rudolf Ernst Peierls
* William George “Bill” Penney
* George Placzek
* Michael J. Poole
* Joseph Rotblat
* Harold Sheard
* Tony Hilton Royle Skyrme (after whom skyrmions are named)
* Geoffrey Ingram Taylor
* Ernest W. Titterton
* James Leslie Tuck

And The #1 British Mission Scientist Linked To Feynman Was…

Klaus Fuchs: when Feynman’s wife was dying of tuberculosis, he borrowed Fuchs’ car to drive to her side at speed. Yes, Fuchs was a Communist who later admitted giving nuclear secrets to the Russians (and so went to jail). And despite being German, he spent a lot of time working in Edinburgh etc, so almost certainly was ‘Britainized’ to a large degree.

But did he make up the Feynman Challenge Ciphers? I don’t know. There were many other bachelors living in the Big House at Los Alamos: Fuchs and Feynman were just two.

Perhaps hints towards the answer will lie in one of the many autobiographies from the people involved, such as “Bird of Passage: Recollections of a Physicist” (Rudolf Peierls), or “What Little I Remember” (Otto Frisch): or indeed in Ferenc Morton Szasz’s British Scientists and the Manhattan Project: The Los Alamos Years.

If there’s a reasonable chance Feynman Cipher #2 (“F2”) is an exotic transposition cipher, it struck me that it might be a good idea to apply a whole load of exotic transpositions, and then use a really simple test to try to order them according to some minimized metric.

The metric I chose was the number of unique letter pairs appearing in the transposed cipher, simply because even a Vigenere should respond to that (for a reasonable-sized cipher). So… here is the bit of C code I wrote:-


#include <stdlib.h>
#include <stdio.h>
#include <memory.h>

const char F2[] = /* Note: 261 = 9 * 29 */
"XUKEXWSLZJUAXUNKIGWFSOZRAWURO"
"RKXAOSLHROBXBTKCMUWDVPTFBLMKE"
"FVWMUXTVTWUIDDJVZKBRMCWOIWYDX"
"MLUFPVSHAGSVWUFWORCWUIDUJCNVT"
"TBERTUNOJUZHVTWKORSVRZSVVFSQX"
"OCMUWPYTRLGBMCYPOJCLRIYTVFCCM"
"UWUFPOXCNMCIWMSKPXEDLYIQKDJWI"
"WCJUMVRCJUMVRKXWURKPSEEIWZVXU"
"LEIOETOOFWKBIUXPXUGOWLFPWUSCH";

#define WIDTH 29
#define HEIGHT 9

int count_unique_pairs(const char *cipher, const int * order)
{
int i, j, n;
int curr, last;
int count[26][26];

memset(count, 0, sizeof(count));

n = 0;
curr = -1;
for (i = 0; i < WIDTH; i++)
{
for (j = 0; j < HEIGHT; j++)
{
last = curr;
curr = cipher[order[j]*WIDTH+i] - 'A';
if ((last >= 0) && (count[last][curr]++ == 0))
n++;
}
}
return n;
}

int best_pair_count = 10000000;

void find_order_with_least_unique_pairs(const int * old_order, int index)
{
if (index < HEIGHT) { int new_order[HEIGHT]; int i; for (i = index; i < HEIGHT; i++) { memcpy(new_order, old_order, sizeof(new_order)); new_order[index] = old_order[i]; new_order[i] = old_order[index]; find_order_with_least_unique_pairs(new_order, index + 1); } } else { int count = count_unique_pairs(F2, old_order); #if 0 if (count > 188) return; #else if (count > best_pair_count) return; #endif best_pair_count = count; printf("pairs = %d, order = { %d %d %d %d %d %d %d %d %d }\n", count, old_order[0], old_order[1], old_order[2], old_order[3], old_order[4], old_order[5], old_order[6], old_order[7], old_order[8]); } } const int default_order[9] = { 0, 1, 2, 3, 4, 5, 6, 7, 8 }; int main(int argc, char argv[]) { find_order_with_least_unique_pairs(default_order, 0); return 0; }

It didn't reveal a great deal (and pasting it into WordPress lost all the formatting, *sigh*), but I thought I'd post it here anyway. 🙂

For those of you who have had their fill of the last week’s posts on the Somerton Man, here’s a different cipher mystery that doesn’t get aired even 1% as much: the Feynman Ciphers.

The first Feynman Cipher (F1, 380 characters long) turned out to be based on a 5 x 76 transposition path cipher (the plaintext was “WHANTHATAPRILLEWITHHISSHOURESSOOTE”, i.e the start of Chaucer’s Canterbury Tales), but what is a little odd is that nobody seems to have yet made any inroads at all into the other two, though it is often remarked that transposition may well be involved. In that sense, they’re a bit like the d’Agapeyeff challenge cipher, which is also believed to be a multi-stage cipher including one or more transposition stages.

At 261 characters long, the second Feynman Cipher (F2) is a little shorter than F1: this length factorizes to 3 x 3 x 29, or 9 x 29, or 3 x 87. It also includes all 26 letters, which rules out a lot of tricky ciphers such as Playfair and Phillips.

XUKEXWSLZJUAXUNKIGWFSOZRAWURORKXAOS
LHROBXBTKCMUWDVPTFBLMKEFVWMUXTVTWUI
DDJVZKBRMCWOIWYDXMLUFPVSHAGSVWUFWOR
CWUIDUJCNVTTBERTUNOJUZHVTWKORSVRZSV
VFSQXOCMUWPYTRLGBMCYPOJCLRIYTVFCCMU
WUFPOXCNMCIWMSKPXEDLYIQKDJWIWCJUMVR
CJUMVRKXWURKPSEEIWZVXULEIOETOOFWKBI
UXPXUGOWLFPWUSCH

Though normally very good at identifying cipher types, Cryptocrack doesn’t do particularly well in this: it suggests Phillips, FracMorse, Playfair and Beaufort as its top four tips, none of which seem hugely likely to me. What is interesting, though, is that if you transpose the ciphertext (say, using some of the seven transposed routes listed by James Lyons), Cryptocrack produces a quite different set of recommendations, suggesting instead Trifid (which it almost certainly isn’t), but more reasonably Running Key and occasionally Vigenere.

Personally, I don’t think it’s a Vig: so right now, my prediction is that it’ll turn out to be a funky path transposition combined with Running Key (combining this with Vigenere would surely be just a bit too sadistic). Perhaps this will be what James Lyons will say too, when he gets round to posting part 3 (his part 2 is here.

Finally: the third Feynman Cipher (F3) is short too: 231 characters, which factorizes to 3 x 7 x 11. Much as James Lyons notes, I currently expect more or less everything said about F2 to hold true for F3: so I epxect it’s probably a Running Key (or perhaps Vigenere, but I doubt it) combined with a funky path transposition.

WURVFXGJYTHEIZXSQXOBGSVRUDOOJXATBKT
ARVIXPYTMYABMVUFXPXKUJVPLSDVTGNGOSI
GLWURPKFCVGELLRNNGLPYTFVTPXAJOSCWRO
DORWNWSICLFKEMOTGJYCRRAOJVNTODVMNSQ
IVICRBICRUDCSKXYPDMDROJUZICRVFWXIFP
XIVVIEPYTDOIAVRBOOXWRAKPSZXTZKVROSW
CRCFVEESOLWKTOBXAUXVB

What do you think?

After I recently mentioned Bellaso’s set of seven challenge ciphers from 1564 on this blog, Augusto Buonafalce very kindly emailed me with scans of Bellaso’s three challenge ciphers from 1555. I’ve now transcribed these (as best I can) and have added them to the existing Bellaso cipher transcriptions page.

I do acknowledge that the font that my theme currently uses for “preformatted” text is too small (thanks Dennis!), but the ciphertexts are only really there to be cut-and-pasted into whatever hacky cryptanalysis package you choose. Incidentally, one neat little online crypto cipher package is John’s Javascript Secret-Code Systems webpage, which has a number of unsolved ciphertexts, such as the three “Richard Feynman” challenge ciphertexts (copied onto a Cipher Mysteries page).