“Both autokeys again, with the tableau built from the puzzle's own words”
0002 and 0012 used the standard and KRYPTOS alphabets only. Here the plain and cipher alphabets are the keyword mixings of every vocabulary word, with the key letters read in the standard, KRYPTOS or the same mixed alphabet.
An autokey cipheran autokey cipher A cipher that uses the message itself, or the scrambled text itself, as the secret word, each letter enciphered with the letter a fixed distance before it. uses the messagethe message The 97 letters of English that Jim Sanborn wrote and enciphered. itself, or the scrambled textthe carved letters The 97 letters cut into the copper at the CIA, shown at the top of every experiment page. itself, as the secret wordthe secret word The word or string a cipher needs besides the method.. We ruled that outdead Said of a period, grid or method when the known letters contradict it: no secret word of that kind can produce the carved letters. yesterday with the sculpture's two alphabetsthe sculpture's alphabet The alphabet reordered to begin with KRYPTOS: KRYPTOSABCDEFGHIJLMNQUVWXZ.. But the tablethe table The right-hand half of the sculpture, 26 rows of the alphabet each shifted one place further, starting from the word KRYPTOS. could have been built from another word, which changes every sum. So we rebuild the table from each word the puzzle knows and run both autokey checks again: for the scrambled-text version the 24 known letters decide at once; for the message version, each known letter reveals another, and the chain either breaks or fills the whole message.
Words in the margin are defined on the glossary page.
1.1
The method at work
The carved letters go in at the top; the reading comes out at the bottom.
the two autokeys
In an autokey the message, or the scrambled text, becomes its own key after a short start. Experiments 2 and 3 tried them on the sculpture's table; this one rebuilds the table from every puzzle word.
the strip
For the plaintext autokey, short lags and lags 33 to 49 die under every alphabet pair. The ciphertext autokey never comes within three letters of the known ones at any lag.
the scale
Where a lag survives, the letters it forces read as noise: -6.47 as English, worse than the best on a shuffled copper.
2
Best candidate against the known letters
This experiment produced no candidate: the search ruled everything out before a full text was built. Its product is the bounded negative in section 7.
3
Walkthrough, by hand
Redo the method on a few letters with pencil and paper.
Build the alphabet from ABSCISSA: ABSCIDEFGHJKLMNOPQRTUVWXYZ, and use it for both the message letters and the carved letters, Vigenere.
Ciphertext autokey, lag 10: the key for position 22 is the carved letter at 12, L.
Look up E in the ABSCISSA alphabet, index 5, and L, index 12; add: 17, which is R in that alphabet. The carved letter at 22 is F, so this is a mismatch.
Count mismatches over all known letters past position 10; any mismatch kills this lag for this alphabet.
Plaintext autokey, lag 10: the key for position 32 is the message letter at 22, E, which we know, so position 32's message letter must be A; it is, so no conflict yet; position 42's letter follows from position 32's, and so on down the chain.
the worksheet: the first steps by hand, on squared paper
4
Worked example, generated
The same steps run by the code on the real letters.
alphabet from ABSCISSA: ABSCIDEFGHJKLMNOPQRTUVWXYZ; used for message, carved text and key; Vigenere
ciphertext autokey, lag 10:
pos plain key = carved at pos-10 E+key carved match
22 E L R F no
23 A B B L no
24 S S I R no
64 B L M N no
65 E U A Y no
plaintext autokey, lag 10: chain from the cribs breaks after forcing 24 letters
5
Hypothesis, for a cryptanalyst
Family, key space, the constraint under test, and the clues it reads.
Ciphertext autokey, c[i] = E(p[i], c[i-L]), is decided by the cribs alone: at every crib position past the lag the key letter is a known ciphertext letter, so each (lag, convention) gets up to 24 exact checks. Plaintext autokey, c[i] = E(p[i], p[i-L]), is decided by the chained closure of 0012. Both were dead under the standard and KRYPTOS alphabets (0002, 0003, 0012); the NSA's 1992 'plain-text auto-key' guess (C11a) and Scheidt's keyword alphabets (C5) leave the mixed-alphabet case open. Alphabet pool: standard, KRYPTOS and the keyword mixing of each vocabulary word (63); plain and cipher alphabets range over the pool, the key alphabet over standard, KRYPTOS and the plain alphabet itself; three modes; lags 1-96. Ciphertext autokey: zero mismatches is a survivor. Plaintext autokey: a consistent closure with 60 or more forced letters is scored as English on its non-crib runs against ten shuffled controls, and a fully forced text is submitted. Bounds: complete for this pool and these key-alphabet choices.
6
Prediction
Written before the run; the result decides it.
Ciphertext autokey: no alphabet triple, mode and lag passes all its checks; the best passes about 6 of 24. Plaintext autokey: every lag up to 25 breaks under every triple; closures with 60 or more letters read as noise against the controls. Zero candidates pass.
7
Method
Build the alphabet pool and the key-alphabet choices.
Ciphertext autokey: for each triple, mode and lag, compare the enciphered crib letter with the carved letter at every crib position past the lag; count mismatches.
Plaintext autokey: for each triple, mode and lag, chain forced letters from the cribs to closure; reject on conflict; record forced-letter count.
Score large closures as English on non-crib runs; repeat on ten shuffles.
Submit fully forced texts.
8
Result and bounds
The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.
0candidates returned
0passed every gate
0/24known letters, best
–lowest loss
nonechampion before
3429216 (triple, mode, lag) cases. Ciphertext autokey: best 99 mismatches, 0 survivors. Plaintext autokey: lags dead under every triple [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 33, 34, 35, 36, 37, 38]...; 0 fully forced; best closure -6.469426635667049 vs shuffled -6.157.
What was searched, as the harness recorded it.
family
ciphertext and plaintext autokey with keyword-mixed alphabets