“F13.5: memorable keys scored by English alone: key words, keyword tables, running keys from the named texts, autokey primers and keyword columnars from the source list, through the methods already built, no known letter read; K4 against 100 shuffles, plants per method”
F13.5 (issue #85); reruns 0242, which errored drawing a running-key plant from a named text shorter than 97 letters (plants now draw only from texts longer than 97; the search still reads every text). Every earlier key-source search (0004-0021, 0045-0053, 0109, 0115, 0117, 0154, 0210) decided by the 24 known letters. Here the same memorable keys are read with no known letter: five methods, every key from the source order (program.md: K1-K3 words, the sculpture's texts, the Morse, Carter, then the dictionary), every reading scored exhaustively by quadgrams, the 20 best per method rescored by MIX8m (0233). A: a periodic key word (every puzzle word and every dictionary word of 3-24 letters, every start) on the 24 rules of the sculpture's tables. B: keyword-mixed tables from the puzzle words (one side or both) with a puzzle key word. C: a running key from 0210's named texts, every offset, both directions, on the 24 rules. D: plaintext and ciphertext autokey primed by any puzzle or dictionary word, on the 24 rules. E: a keyword columnar keyed by a puzzle word, before or after the table with a puzzle key word. Statistic: max z over the five methods against K4's 100 shuffles; plants (8 per method, keys drawn from the same sources) each against 10 own shuffles.
The creators said the keythe secret word The word or string a cipher needs besides the method. could be remembered: a word, or a passage of a known text. Earlier runs checked every such key against the clue wordsthe known letters Sanborn has confirmed that positions 22 to 34 of the message read EASTNORTHEAST and positions 64 to 74 read BERLINCLOCK.. Here we check them again without looking at the clue words at all: every key word, every passage of the sculpture's texts and Carter's book, every keyword-scrambled table, read in full and keptthe champion The candidate with the lowest loss so far among those that passed every check. only if the result reads as English. K4the carved letters The 97 letters cut into the copper at the CIA, shown at the top of every experiment page. is compared with shuffles of its own letters treated the same way, and hidden test messages made with keys from the same lists show whether a right key would stand out.
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
F13.5: keys a person could remember (puzzle words, any dictionary word, passages of the named texts) read without the clue words.
the check
Every key read in full and scored as English; the best per way rescored by the strongest model. K4 against 100 shuffles.
the bound
Plants found 40 of 40; K4 inside its shuffles (p 0.099): closed over the stated key lists.
2
Experiment lineage
What this experiment was built on, from the lineage page: 14 experiments before it, back to the root of its line. Its number is in bold.
keptreverted, in its family's colour×error or timeoutreverted under today's rulesbuilt on a parent in the same familybuilt on a parent in another family
3
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?SORD
YKNOXGCWPBJRMAPCHSQMBGISTYSTOTA
RTPWRENOWJOTSTWEINHLQZTHTEZNEOF
WHAROYBDHHOVFFBEVBIRECASOFMAMWE
checked: the known letters in place, and the reading turns back into the copper
4
How the score is made
The loss of the best candidate, part by part. Zero is solved.
Known letters missed
22 of 24, ten points each
220.00
English
the 73 free letters score -6.59; English text scores -3.6
2.99
Key
35 bits to write it down (key words are cheap, arbitrary letters 4.7 bits each); each 60 beyond 60 costs a point
0.00
Loss
of 240, the worst possible
222.99 92.9%
5
Walkthrough, by hand
Redo the method on a few letters with pencil and paper.
Take a key from the list, say the word ABSCISSA from its third letter, and a way of using the sculpture's table.
Decipher all 97 carved letters with it.
Score the result by how often its four-letter groups occur in English.
Do this for every key word and start, every passage of the named texts, every keyword table and every way of using the table; keep the best.
Do the same on shuffles of K4 and compare.
the worksheet: the first steps by hand, on squared paper
6
Worked example, generated
The same steps run by the code on the real letters.
Key: method A, mode var, pa std, ca kry, ka kry, key MEZUZAH, start 4
carved OBKRUOXOGHULBSOLIFBBWFLRVQQPRNGKSSOTWTQSJQSSEKZZWATJKLUDIAWI
reading EPOTGETEUCNCHBEYDETHSLYPOFTYAABSRFADESMRPPFNZSKYSGLESCUFOOLH
Every key in the lists is read like this, in full, and scored by English alone; no known letter is read.
7
Hypothesis, for a cryptanalyst
Family, key space, the constraint under test, and the clues it reads.
C5 ('simple, can be remembered, executed with the correct key word/s'), C6(a) (keys concealed at the sculpture), C7(a) (K1-K3 words as keys), C25(a) (the Morse words), C20(a) (layer two). Each method was closed exactly with the known letters (0005, 0009-0011, 0020, 0053, 0109, 0115, 0117, 0154, 0210; autokeys 0002, 0003, 0025, 0064; keyword columnars 0110-0114). Those verdicts stand only if the 24 letters sit where stated, letter for letter, without slips (A1, A2) and the plaintext is not masked out of their places. F13.5 is that hedge: the key space is at most about 30 bits per method (log2 of 54 million settings), far under what 97 letters of English carry, and every setting is read, so where the known letters are exact this adds nothing; it decides only the cases where they are not. Pre-registered: S = max over the five methods of (best MIX8m minus the shuffles' mean) / their sd; K4 flags at p < 0.05 against 100 shuffles (one test); a plant is found when its S is above all 10 own shuffles' S; a method is decided when at least 6 of its 8 plants are found.
8
Prediction
Written before the run; the result decides it.
Plants found 8 of 8 in every method (the keys are read exhaustively, so the true key's reading is scored directly). K4 inside its shuffles (p above 0.05), its readings junk with at most 3-4 of 24 known letters in place (harness gate 2). F13.5 closes as exact over the stated key lists.
9
Method
Key lists: 0115's 171 puzzle words (every start), the 234,288-word dictionary (lengths 3-24, every start: 2,241,060 rotations), keyword-mixed alphabets of the puzzle words, 0210's named texts both ways.
Every setting decrypted in compiled code and scored by the harness's quadgram table; the 20 best per method kept, rescored by MIX8m.
Texts: K4, 100 shuffles, 40 plants (8 per method, half with X stops, Carter's held-out text, keys drawn from the lists) and 10 shuffles of each.
z per method against K4's shuffles; S = the largest z; K4 ranked; each plant among its own copies.
K4's best reading per method submitted; gate 4's controls are the shuffles' best readings.
10
Result and bounds
The harness's record: candidates, gates, loss, and the negative result this experiment leaves behind.
5candidates returned
0passed every gate
2/24known letters, best
92.9%lowest loss
nonechampion before
All 5 candidates, lowest loss first.
1failed at cribs2/24 known lettersloss 92.9%
ECDMRIPFEIMEHNLSSTTRTVDOHW?AFLU
EJTKWYOROMEGGSSNAMXWRCWAIDANIZP
PPFOLFNUTCREMLTYDUPEADDLEXFEWIK
NOTCHOBIKILIAFNXNMIMPMKFIFAYYFS
2failed at cribs2/24 known lettersloss 92.9%
ECDMRIPFEIMEHNLSSTTRTVDOHW?SORD
YKNOXGCWPBJRMAPCHSQMBGISTYSTOTA
RTPWRENOWJOTSTWEINHLQZTHTEZNEOF
WHAROYBDHHOVFFBEVBIRECASOFMAMWE
3failed at cribs1/24 known lettersloss 97.0%
ECDMRIPFEIMEHNLSSTTRTVDOHW?EPOT
GETEUCNCHBEYDETHSLYPOFTYAABSRFA
DESMRPPFNZSKYSGLESCUFOOLHELDSJR
OPDNYTRSIMFZGMWBIFOUONZUCUSOBSA
4failed at cribs1/24 known lettersloss 97.0%
ECDMRIPFEIMEHNLSSTTRTVDOHW?LRES
BAMVEGLMVCDHFGHOFTESWTFUZAKPHED
KEAPOSSIXLYWEBOYFARWRFHEINEMQIU
VZHURHOJYEVJMQNESHOCKVHKSCUDIEU
5failed at cribs1/24 known lettersloss 97.3%
ECDMRIPFEIMEHNLSSTTRTVDOHW?RDAL
SRDRIONBLURSJPFLCUSLIXPACBTAEJS
DWMEEXBJCKARSCFPSABEZBOFJDUVIAD
IODOSHDAZVZZZGMNNZBIANMELEEARAC
K4 statistic 2.23 (method C), p 0.099 against 100 shuffles; plant power A 8/8, B 8/8, C 8/8, D 8/8, E 8/8; decided ['A', 'B', 'C', 'D', 'E']
What was searched, as the harness recorded it.
item
F13.5
search reads known letters
no
methods
A
24 rules x (1,084 puzzle-word starts + 2,241,060 dictionary rotations)
B
5,100 keyword-table setups (170 puzzle alphabets) x 1,084 puzzle-word starts
C
24 rules x 36 named texts (both directions), every offset: 596,530 offsets
D
24 rules x 2 kinds x 233,014 primer words
E
138 keyword columnars x 2 orders x 24 rules x 1,084 puzzle-word starts
exhaustive
yes
scoring
every setting read in full and scored by the harness quadgram table; the 20 best per method rescored by MIX8m (0233)