“F1.1 decided with F1.5's design 2: K4 against 12 scrambled copies of itself at key length 23 (one free alphabet, Beaufort form); design 2's power at length 26”
F1.5 step 3 and F1.1. Design 2 (0139) noticed planted English against shuffles of itself 6 times in 8. Here it runs on K4 and on 12 shuffles of K4 that fit length 23 exactly, the like-for-like comparison 0138 showed is needed, so F1.1's Beaufort-23 cell gets a verdict at stated power. In the same run, design 2's power at length 26 (one check) is measured as in 0139: 6 planted texts, each against 6 fitting shuffles of itself.
We finally have a search that, on planted messages, tells hidden Englisha planted message A real English passage, with the 24 known letters written in at their places, enciphered by the method under test with a random key. from its own scrambled letters about three times in four for one kind of cipher (one scrambled alphabet, a 23-letter key, used the BeaufortVigenere The three rules of arithmetic a table can use: add the key letter to the message letter, subtract the message letter from the key letter, or subtract the key letter from the message letter. way). Now it runs on the carved textthe carved letters The 97 letters cut into the copper at the CIA, shown at the top of every experiment page. itself, against scrambled copies of the carved text. If the carved text were this kind of cipher, the search would usually rank it first. We also check whether the search works for a 26-letter key, where the known lettersthe known letters Sanborn has confirmed that positions 22 to 34 of the message read EASTNORTHEAST and positions 64 to 74 read BERLINCLOCK. give only one check.
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 idea
Design 2 notices planted English against its own scrambled letters about three times in four at length 23. Now K4 runs against scrambled copies of K4, the same like-for-like comparison.
the check
12 shuffles of K4 that still fit the known letters at length 23; the same search on all 13. K4 scores -4.99, copies -5.14 to -4.83.
the bound
K4 ranks 9 of 13: the Beaufort-23 cell is closed at the measured strength. At length 26 design 2 notices 5 of 6 planted texts, so K4 at 26 is next.
2
Best candidate against the known letters
Laid out as the copper carries K4: OBKR ends the line that finishes K3, then three lines of 31. Left, as carved, the 24 known places in accent. Right, the candidate: a known place it gets right is an accent cell, one it misses is outlined. This is the candidate with most known letters right; none passed every gate.
in: the copper, always the same
as carved, 97 letters; the 24 known places in accent
ECDMRIPFEIMEHNLSSTTRTVDOHW?OBKR
UOXOGHULBSOLIFBBWFLRVQQPRNGKSSO
TWTQSJQSSEKZZWATJKLUDIAWINFBNYP
VTTMZFPKWGDKZXTJCDIGKUHUAUEKCAR
OBKR ends the line that finishes K3, then three lines of 31
out: the candidate
every place read; known places in accent, misses outlined
ECDMRIPFEIMEHNLSSTTRTVDOHW?BOAR
IMELRBCEHINVEHWATEASTNORTHEASTH
ATEGNANHIFPLLBADIRODMOROSORYBER
LINCLOCKDAVARRUNDISHEDSAVEANLAP
checked: the known letters in place, and the reading turns back into the copper
3
How the score is made
The loss of the best candidate, part by part. Zero is solved.
Known letters missed
0 of 24, ten points each
0.00
English
the 73 free letters score -5.28; English text scores -3.6
1.68
Key
198 bits to write it down (key words are cheap, arbitrary letters 4.7 bits each); each 60 beyond 60 costs a point
2.31
Loss
of 240, the worst possible
3.98 1.7%
4
Walkthrough, by hand
Redo the method on a few letters with pencil and paper.
Write K4 in rows of 23; each column is one key slot; seven columns hold two known letters.
Run sixteen searches side by side over alphabets that keep those seven pairs consistent, swapping states between bold and cautious searches, and keep the best English score.
Shuffle K4's letters into 12 new orders that could still give the 24 known letters, and run the same search on each.
If K4 were English enciphered this way, it would usually come first; if it comes anywhere inside the pack, this cell is closed at the measured strength.
Do the same check of strength at 26 letters on planted English.
the worksheet: the first steps by hand, on squared paper
5
Worked example, generated
The same steps run by the code on the real letters.
key length 23, Beaufort form: K4 in rows of 23; each column is one key slot
O B K R U O X O G H U L B S O L I F B B W F L
R V Q Q P R N G K S S O T W T Q S J Q S S E K
Z Z W A T J K L U D I A W I N F B N Y P V T T
M Z F P K W G D K Z X T J C D I G K U H U A U
E K C A R
the seven checks every move keeps (place(carved) + place(known) equal within a column):
column 0: R+S (24) = M+C (70)
column 1: V+T (25) = Z+L (71)
column 2: Q+N (26) = F+O (72)
column 3: Q+O (27) = P+C (73)
column 4: P+R (28) = K+K (74)
column 21: F+E (22) = T+I (68)
column 22: L+A (23) = T+N (69)
best alphabet on K4: RTKDPNVSXJQMUAZGOBILEYCHWF
key values: [7, 7, 15, 0, 4, 1, 2, 9, 15, 14, 8, 13, 14, 25, 21, 25, 12, 22, 15, 4, 25, 19, 6]
place 22: carved F at 25, key 19: 19 - 25 = 20 -> E
place 64: carved N at 5, key 22: 22 - 5 = 17 -> B
6
Hypothesis, for a cryptanalyst
Family, key space, the constraint under test, and the clues it reads.
C27(c) (a tableau 'slightly different from the one on Kryptos', clue register), C1(b) (the table alone). Model as 0137: one unknown alphabet on both sides, Beaufort form x(c_i) + x(p_i) = k[i mod L], L 23 (7 checks) and 26 (1 check), the only (form, length) pairs 0119 left. Search: 0139's compiled parallel tempering in the set where every check holds (10 runs x 16 replicas x 625 rounds x 1,000 moves, T 0.1 -> 0.002), full-text quadgram mean. L23: K4 and 12 shuffles of K4 that fit exactly (0119's decider, seed 140). Measured power (0139): a planted Beaufort-23 text beat all 6 of its own shuffles 6 times in 8; against 12 the power is lower, about 0.5-0.75. Verdict rule fixed before the run: K4 above all 12 of its copies stops the queue for a replication with fresh copies; otherwise the cell is closed, sampled at the stated power (false alarm rate 1/13). L26 power: 6 plants (seed 1400) x 6 own fitting shuffles; bar 5 of 6; if met, K4 at 26 is the next run, otherwise the 26 cell stays blocked.
7
Prediction
Written before the run; the result decides it.
K4 is not above all 12 of its copies at length 23 (inside the range), so F1.1 Beaufort 23 closes, sampled at power about 0.6. At 26, fewer than 5 of 6 planted texts beat all their own copies (one check gives the climb little to hold), so the 26 cell stays blocked. No reading. One candidate (K4's best at 23) carried to the gates with its copies as controls; it will not pass.
8
Method
Draw 12 shuffles of K4 that fit length 23 exactly; draw 6 planted texts at length 26 and 6 fitting shuffles of each.
Run 0139's search on all 55 texts at equal effort.
L23: K4's rank among its 12 copies. L26: plants beating all their own copies.
Return K4's best L23 reading with the copies' English scores (free letters) as controls.
9
Result and bounds
The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.
1candidates returned
0passed every gate
24/24known letters, best
1.7%lowest loss
nonechampion before
All 1 candidates, lowest loss first.
1failed at baseline24/24 known lettersloss 1.7%
ECDMRIPFEIMEHNLSSTTRTVDOHW?BOAR
IMELRBCEHINVEHWATEASTNORTHEASTH
ATEGNANHIFPLLBADIRODMOROSORYBER
LINCLOCKDAVARRUNDISHEDSAVEANLAP
Beaufort form, one alphabet RTKDPNVSXJQMUAZGOBILEYCHWF on both sides, key length 23
L23: K4 -4.989, rank 9 of 13 among its own copies (-5.137 to -4.833); stop queue False. L26 power: 5/6 plants beat all 6 own copies (null 0.86, bar 5).
What was searched, as the harness recorded it.
family
one unknown alphabet on both sides, Beaufort form, no move: K4 at key length 23 against its own copies; design 2's power at 26
search
parallel tempering (0139), 10 runs x 16 replicas x 625 rounds x 1000 moves per text, T 0.1->0.002