“A6.3: masked messages (reversed, an X every kth letter, the message in every nth place) planted through F1.5's design 2 at one free alphabet, Beaufort length 23, found by the decider, by the mask's own battery test and by the judge on the unmasked reading”
A6.3 (Stage 2, step 4, issue #84). A13 found that a forward English climb overwrites a masked message (1 of 22 masked plants survived 0034/0140/0142's climbs, 0216-0217). Here the question is put to F1.5's design 2, the decider behind F1.1 and F1.2's sampled closes: the three masks A13.1's battery sees well (reversal: eng_reverse 20 of 20; an X every kth letter: eng_drop_nth 20 of 20; the message in every nth place, random letters between: eng_keep_nth 20 of 20) are applied to Carter, the 24 known letters set, enciphered at F1.1's Beaufort length 23 and climbed by 0140's frozen search with six fitting shuffles of each plant's own ciphertext.
Scheidt said he disguised the Englishthe message The 97 letters of English that Jim Sanborn wrote and enciphered. before enciphering it. If the message was written backwardsa mirror Reading or writing the text from the end to the start., or had an X every few letters, or was spread out with junk letters between, a search that tunes the keythe secret word The word or string a cipher needs besides the method. toward ordinary English may walk straight past the right key. Here we hide disguised messages, run the same search on them, and check three ways of noticing them: the search's own score, a pattern test made for that disguisea mask Something done to the English before the cipher proper, such as writing it backwards or slipping in meaningless letters, that is not itself a cipher but moves or hides the letters., and the judgethe judge A language model that scores how much of a reading is English, from 0 to 100, allowing misspellings and X between sentences. reading the text after the disguise is undone.
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
Scheidt masked the English before enciphering it. If the mask leaves no forward English, a climb tuned to English may pass the right key by.
the check
Carter's prose reversed, with an X every kth letter, or spread one letter in every nth place, enciphered at F1.1's length 23 and climbed by 0140's frozen search beside six scrambled copies each.
the bound
Reversed messages are found by the climb's own score, 7 of 8. An X every kth letter (best 5 of 8) and the message in every nth place (best 3 of 8) are not: F1.1 and F1.2's sampled closes hold for unmasked or reversed English only.
2
Experiment lineage
What this experiment was built on, from the lineage page: 15 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
This experiment produced no candidate: the search ruled everything out before a full text was built. Its product is the bounded negative under Result and bounds below.
4
Walkthrough, by hand
Redo the method on a few letters with pencil and paper.
Take a sentence from Carter and write it backwards.
Put EASTNORTHEAST and BERLINCLOCK at their places.
Encipher it with a shuffled alphabet and a key of 23 numbers in the Beaufort form.
Make six scrambled copies that the same key length still fits, and run the same search on all seven.
Undo the disguise (read the reading backwards) on all seven and ask the judge which reads as English.
The disguised message is found if it beats all six copies.
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.
A6.3 plants a masked message. A reversed Carter passage with the known words set after masking:
plant TIYROTSEHTSIFEIRBNITAEASTNORTHEASTDNAYRUJREPROFDEIRTSAWEHTINIDNBERLINCLOCKIHDNUOBSAWEGDUJASANAMYP
unmasked PYMANASAJUDGEWASBOUND?????????????HI???????????NDINITHEWASTRIED???????????FORPERJURYAND?????????????ATINBRIEFISTHESTORYIT
enciphered at one free alphabet, Beaufort, key length 23; climbed by 0140's search with six copies;
found by the climb's score, by eng_reverse, or by the judge on the reading read backwards.
6
Hypothesis, for a cryptanalyst
Family, key space, the constraint under test, and the clues it reads.
A6 and A13 (knowledge/12-assumptions.md); C4(a) (Scheidt 2005, 2015: 'I masked the English language', a plaintext transform before the keyed stage; 'more than one stage'), C19 (X as a stop). Method under test: F1.5 design 2 (parallel tempering in the check-keeping set, 10 runs x 16 replicas x 625 rounds x 1000 moves) on F1.1's Beaufort form, one alphabet, length 23 (7 checks), whose close (0139-0140) rests on design 2 finding 6 of 8 clean plants above their own copies. A masked plaintext has the right key and the right alphabet but its free letters are not forward English, so the climb, scored by forward quadgrams, has no reason to settle on them. Three ways to find a masked plant, each against its own 6 copies (null 1/7): (a) the decider's own rule, the climb's quadgram score; (b) the mask's own test from tools/a13_structure.py on the reading (eng_reverse, eng_drop_nth, eng_keep_nth); (c) the judge on the reading unmasked with the planted mask's own parameters (an oracle unmask, applied to the copies' readings alike, the 24 known places set to '?' before unmasking so the judge sees no known word), one blind call. Pre-registered: if no way finds at least 6 of 8 plants of a mask (0139's bar), the sampled closes of F1.1 and F1.2 stand as 'closed for unmasked English' only, for that mask.
7
Prediction
Written before the run; the result decides it.
The decider's own rule finds 0-2 of 8 per mask; the mask's own test finds at most 2 of 8 (the climb fills the free letters with forward junk, as in 0216); the judge on the unmasked reading finds at most 2 of 8, because the climbed readings get about a third of the letters right. So F1.1 and F1.2's sampled closes hold for unmasked English only. Zero candidates.
8
Method
Per mask, 8 plants: Carter (from place 20,000 on, random start), masked as tools/a13_check.py's power test masks it (reversal; X at every kth place, k 5-11; the message at every nth place from offset o, n 2-4, random letters between), the 24 known letters set after masking.
Enciphered with a random alphabet and key of 23 in the Beaufort form (0120's rule); 6 own copies, shuffles of the ciphertext that 0120's exact decider fits at 23.
Every text through 0140's frozen _job (design 2), all runs to the end.
(a) climb score above all 6 own copies'; (b) the mask's battery test above all 6; (c) one judge call on every reading unmasked with the plant's parameters, known places '?', spacers inserted at the judge's withheld places 22-34 and 64-74 (it masks those places of any text) so no unmasked letter is hidden; above all 6.
9
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
reversal: decider 7/8, test 2/8, judge 0/8; X every kth: decider 5/8, test 3/8, judge 0/8; message in every nth: decider 1/8, test 1/8, judge 3/8
NIWYROEIIEMORPPAEDEDS ### IE ??? OTAFIREVEDRTNIVY ### SEMANIOWECARF ??? MORSILGNANENOLEMIGENE (no coherent words in forward reading; reversed gives only fragments such as APPRO, NAMES, ENE, ONE)