## Saturday, 25 January 2014

### A look at the second Feynman cipher [part 2]

The Feynman ciphers are a set of 3 ciphers given to Richard Feynman, the first of which has been solved, but the second 2 remain unsolved. This is part 2 of a discussion about the second cipher, which I refer to as F2.

In the previous part, I was playing around with vigenere type ciphers, in this part I have decided to step back a little bit a figure out what the ciphertext is not i.e. identify the cipher algorithms it can't be. To do this we need a comprehensive list of cipher algorithms, thankfully there are lists around.

Using the list of ciphers above plus a few more from here, I have created the following list of ciphers, ruling out the ones it can't possibly be:

Transposition: AMSCO, COMPLETE COLUMNAR TRANSPOSITION, INCOMPLETE COLUMNAR, MYSZKOWSKI, NIHILIST TRANSPOSITION, RAILFENCE, ROUTE TRANSPOSITION, REDEFENCE, SWAGMAN

Substitution: SIMPLE SUBSTITUTION, BACONIAN, CAESAR, ROT13, AFFINE, ATBASH

Uses 25 letter alphabet: BAZERIES, BIFID, CADENUS, CHECKERBOARD, CM BIFID, FOURSQUARE, PHILLIPS, PHILLIPS-C, PHILLIPS-RC, PLAYFAIR, SERIATED PLAYFAIR, TRI-SQUARE, TWIN BIFID, TWO-SQUARE

More than 26 chars: DIGRAFID, HOMOPHONIC, TRIFID, TWIN TRIFID

Other: GRANDPRE (ct is numbers), GRILLE (ct length must be square number), GROMARK (ct starts with 5 digits), PERIODIC GROMARK (ct starts with digits), HEADLINES (ct has spaces), KEY PHRASE (word divisions retained), MONOME-DINOME (ct is numbers), MORBIT (ct is numbers), NIHILIST SUBSTITUTION (ct is numbers), POLLUX (ct is numbers), RAGBABY (word divisions are kept, 24 chars), TRIDIGITAL (ct is numbers), ADFGX/ADFGVX (only 5/6 characters used), POLYBIUS SQUARE (only 5/6 characters used), HILL (2*2)(requires even number of characters)

The above ciphers can be ruled out. The transposition ciphers because the frequency distribution does not match english, the substitution ciphers because the IC is 0.045, which is too low for English. The F2 cipher also has 26 characters in total, so it can't be anything based on a 5 by 5 key square. Some ciphers have too many characters (more than 26) so they can also be ruled out. Others can be ruled out because the ciphertext (ct) they produce consists of numbers.

# The Possibilities

The following are the ciphers I can't rule out: FRACTIONATED MORSE, GRONSFELD, AUTOKEY, INTERRUPTED KEY, NICODEMUS, NULL, PORTA, QUAGMIRE I, QUAGMIRE II, QUAGMIRE III, QUAGMIRE IV, RUNNING KEY, VARIANT, VIGENÈRE, PORTAX, PROGRESSIVE KEY, SLIDEFAIR.

Assumptions: we all have to make assumptions, it is better to lay them out at the start. 1. The plaintext is in English, 2. There are no NULLs distributed through the ciphertext, 3. It is not a combination transposition + something else. 4. When making the ciphertext, no mistakes were made that prevent decryption.

I may have to weaken some of these assumptions later, but for now I think focusing on simpler stuff should be prioritised.

# Narrowing Possibilities

In part 1, I tried solving it using Vigenere, Beaufort, Variant Beaufort, Gronsfeld, VigAutokey and Porta, and this was unsuccessful which makes me confident I can ignore these possibilities under the assumptions above.

I have used this site to get a list of candidate ciphers ranked by likelyhood. I have also removed the ones I know are not possible:

FracMorse 5
RunningKey 8
Quagmire3 8
Quagmire2 9
Quagmire4 9
Nicodemus 11
Vigslidefair 12
Portax 13
Swagman 22
Progkey beaufort 23
Progressivekey 24


This means I will concentrate my focus on FracMorse, Running Key and the Quagmires. I will update again in Part 3.