Go ahead and post it here.
I am not sure what territory you guys have covered. But I see a lot of pictures in the cryptography books, and sometimes people make route ciphers by breaking the message down into more than one smaller shapes. Have you guys covered this yet? I was thinking that the more "cuts" there are, meaning the more shapes there are, the more difficult it would be to make comparable period 15/19 repeat stats. That’s because there are fewer contiguous columns to create the repeats. And the fewer "cuts" there are, meaning the fewer shapes there are, the easier it would be to make comparable period 15/19 stats. It would be an interesting experiment to find out what is possible.
Have you guys tried anything something similar to this ( beginning of message shown to demonstrate inscription direction )?
When you pixilate the code you see this patern developing.
Could this patern have markers of some sort in it?
Two other points, that are not related to each other but may possibly be shown with the mirrored period x conditionally formatted table:
1. See the pink boxed periods. They are darker and generally have higher counts except for the period 15 and period 19. How could this be if the inscription rectangle was 15 x 22? But if Zodiac used multiple inscription rectangles of relatively narrow width, that may explain the high count of low period repeats.
The first two letters of the message below are I and L. If transcription is vertical, then they change from period 1 to period 15. The first and third letters of the message are I and I. they would change from period 2 to period 30. But the first and fourth letters are I and K, and they change from period 3 to period 1. There should be a fairly high count of period 3 matches, because of repeating words or word fragments. This would show up as period 1. Not as many as period 15 or period 30, but still relatively high as compared to the other periods.
2. I was also thinking about making an experiment to skip a varying count of plaintext during transcription for a lot of randomly created messages. To see how easy it is or how many symbols I need to skip to create repeat distortion zones and reproduce the period 15/29 phenomenon ( boxed yellow ).
What do you guys think? Should I work on multiple inscription rectangles, or the skips? Or did you already work on multiple inscription rectangles?
EDIT: Jarlve, you can make me a diagonal transposition message if you want. I don’t know if I will be able to figure out direction, but hoping to get a decent solve score.
EDIT 2: Is there a way to make inscription rectangles so that period 1 becomes period 19, period 2 becomes period 39 ( caused by nulls ), and period 3 becomes period 5? With the mirrored version, there are other high count periods boxed pink. Is there a way to make inscription rectangles so that period 1 becomes period 15, period 2 becomes period 29 ( with skips, which to me seem possibly more likely because they could be passive mistakes instead of an action like adding nulls ), and period 3 or 4 become period 1, 2, 3, 4, 5 or 6? That is the model that I am thinking of.
Could this patern have markers of some sort in it?
How much ink was used? 
Have you guys covered this yet? I was thinking that the more "cuts" there are, meaning the more shapes there are, the more difficult it would be to make comparable period 15/19 repeat stats.
I did a quick test that considers any such part size , untransposing each part with the period operation, some weeks ago. It’s something that we could look into. But have my hands full with *just* columnar transposition at the moment.
1. See the pink boxed periods. They are darker and generally have higher counts except for the period 15 and period 19. How could this be if the inscription rectangle was 15 x 22? But if Zodiac used multiple inscription rectangles of relatively narrow width, that may explain the high count of low period repeats.
The periods at which the encoding is actual – if sequential – are typically a bit higher in repeats. But I’m not sure if that is enough of an explanation in the case of the 340.
2. I was also thinking about making an experiment to skip a varying count of plaintext during transcription for a lot of randomly created messages. To see how easy it is or how many symbols I need to skip to create repeat distortion zones and reproduce the period 15/29 phenomenon ( boxed yellow ).
If I remember correctly one of your ciphers has the same going on, somewhere between smokie9 and smokie17 I think.
EDIT: Jarlve, you can make me a diagonal transposition message if you want. I don’t know if I will be able to figure out direction, but hoping to get a decent solve score.
Okay, I’ll make one. I was also hoping to make you a couple of ciphers where you try to distinguish between columnar and diagonal.
Smokie, I have not been able to solve your ICT cipher yet. I don’t understand why my own ICT cipher solved qo quickly, just lucky I guess. I’m trying a couple of things, that may take time. The length of your cipher without filler is 319, the filler was added after transposition, and at the end of the cipher (bottom)?
Diagonal cipher:
pr1_jarlve3 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 2 9 25 26 27 11 28 29 30 9 3 31 32 17 33 1 3 18 6 10 8 15 4 9 21 32 34 35 13 33 14 36 19 5 37 7 1 22 38 28 39 20 26 36 5 31 10 7 23 12 40 26 41 42 43 19 16 22 4 9 21 29 24 30 44 39 36 2 28 31 34 45 7 37 5 10 46 20 44 25 47 48 35 7 15 32 8 29 49 19 16 40 15 33 38 46 1 42 12 7 4 22 9 13 39 16 47 6 30 9 21 50 51 15 28 35 27 16 26 11 50 14 37 7 12 31 10 20 52 53 32 33 43 7 7 1 52 3 51 15 19 51 18 22 39 53 28 43 23 29 31 36 5 17 2 45 4 41 52 12 7 16 30 15 10 53 16 37 21 32 33 51 19 11 25 1 7 43 7 26 27 42 8 36 20 44 47 13 54 5 7 3 34 18 23 17 22 26 24 7 4 29 46 35 3 28 30 21 6 2 32 31 7 11 10 36 33 19 18 5 37 22 7 14 38 42 23 7 2 11 44 17 26 45 7 46 52 18 40 47 1 36 15 28 7 23 53 20 2 16 31 15 16 39 7 45 29 41 43 30 24 50 25 37 34 7 35 10 12 8 20 9 19 9 15 4 21 32 33 22 27 29 48 28 51 31 51 10 1 40 39 38 13 9 11 30 18 14 37 54 6 42 52 4 21 48 12 54 32 45 19 22 48 23 53 45
I just found out what my trouble with smokie’s latest ICT cipher is. I’m messing up transposition with untransposition. My solved test cipher transposed the message into random (by key) columns, while actual CT is the other way around. It is transposition by columns, not into columns. Will now try to find out if smokie transposed from top-to-bottom or from bottom-to-top.
Yes, just take out the last 21 symbols. I shouldn’t have put them in there. I used an irregular columnar transposition cipher as Lasry described, not a route cipher where contiguous columns are transcribed creating a lot of period 19 repeats. The last four columns are only 18 rows. I transcribed the columns in alphabetical order according to the key letters, where if one or more key letter were the same I transcribed the leftmost first, working my way to the right.
The 18 row columns cause misalignments in the plaintext, just as we have been working with for our route transposition. Let me know if you want the key, the plaintext version before homophonic encoding, or a complete columnar transposition message with the same plaintext.
Okay thanks for confirming that once more.
Judging from the average scores from my solver for corners 1, 2, 3 and 4 I could say that it looks like you transcribed from bottom-to-top, is that correct?
Top to bottom.
I re-combined enough of the repeats to get a "decent" solve score and a lot of long words and sentence fragments.
Score: 22051 Ioc: 678 M: 167 C: 323 S: 54 standyourrochsofc listhagirltherest hehartofitisthatw illsenadaeiwillbe ngdeoerornandarad dlereddeandallthe auseitshavehilled ssomuchcsllbecome unitismoryslavess ecunthanhisllnotg llingwilddaveyoum meintheforrenamer stbecausemanadaus sthemostdangeryou ousanimalocalltil ohallsomethingdit vesmethemostthril lingenderendestas evenbetterthanget
I re-combined enough of the repeats to get a "decent" solve score and a lot of long words and sentence fragments.
Yes, well done. I was not intending to use the p1 plaintext, it was by accident.
Here’s another diagonal one with a fresh plaintext:
pr1_jarlve4.txt 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 21 32 33 34 35 6 36 30 37 38 4 5 39 8 16 17 40 1 41 14 34 9 13 36 3 42 11 31 43 33 12 20 2 23 7 32 10 35 6 26 40 19 37 44 12 45 1 3 39 17 46 24 36 43 22 30 25 40 42 4 13 38 18 31 47 2 48 23 7 28 37 15 39 17 21 44 9 10 33 5 24 16 25 26 20 34 4 35 29 48 2 1 7 5 23 32 37 8 6 45 18 11 10 47 46 39 15 41 44 13 24 22 25 28 31 33 11 16 45 2 17 35 27 22 14 30 43 41 7 23 18 37 20 26 49 39 19 32 41 48 4 10 26 1 5 38 27 29 14 17 13 47 6 9 31 23 12 16 30 37 44 18 32 28 33 39 44 24 19 35 3 27 1 18 15 29 11 47 29 13 46 25 8 17 43 2 14 20 47 6 31 33 30 9 46 50 32 4 22 7 10 34 5 21 6 15 21 35 9 28 24 11 22 43 23 8 38 19 44 25 1 13 28 27 37 12 34 45 39 11 38 2 42 31 36 34 18 46 41 30 33 29 38 17 32 34 44 47 48 16 35 38 6 22 40 15 21 45 1 48 9 28 11 7 20 49 4 34 15 12 13 14 19 10 18 38 31 5 12 30 43 33 32 27 24 23 25 16 22 46 6 48 30 14 3 2 32 37 28 12 39 11 17 7 20 35 9 29 1
I think that I am going to pursue multiple shapes in the near future. My un-transpositions are crude. I just re-connect the period x symbols into period 1 and get a decent solve score without even knowing the exact inscription and transcription scheme. When we do that with the 340, it doesn’t work. So there is something else going on.
One possibility is multiple inscription rectangles, which limit the length of word fragments that can be recovered to the width of the rectangles. Untransposing as one big rectangle would re-connect the period x symbols to some extent, but there may not be fragments long enough for the solver to pick up on.
On the other hand, the number of inscription rectangles limits the number of possible period 15/19 repeats because of the "cuts" between the rectangles. Fewer contiguous rows = fewer period 15/19 repeats. And even with a 63 symbol key that does not diffuse the plaintext efficiently ( only a few symbols mapping to high frequency plaintext ), it is not particularly easy to create an entirely homophonic message with so many period 15/19 repeats.
I will make a multiple inscription rectangle route transposition cipher message soon to experiment.
EDIT: Thanks for the new message.
.
Here are the cycle stats for smokie18e:smokie18e 51224.
340: 63400.
Smokie,
I’m putting together a compilation of stats for all the test ciphers I’ve encountered in these forums and elsewhere, in an effort to measure which ones are closest to the original 340’s stats with respect to n-grams, repeating fragments, pivots, periods, cycles, etc.
How do you compute your above cycle stats? I’m curious about adding it to my list of cycle-related measurements. I have a few measurements that take different approaches, including Jarlve’s "m_2s_cycles" measurement.
I will look at my exact formula tonight and get back to you.
EDIT:
I find all of the two symbol relationships and score each cycle within that relationship and add up the scores.
For instance, symbols 11 and 36. Red is 2 ^ 8, with a score of 256. Blue is 2 ^ 5, with a score of 32. Green is 2 ^ 2, with a score of 4. Purple is 2 ^ 1 with score of 2. I add up the scores and get 294 for symbols 11 and 36.
At some point I decided to not include the first symbol in any cycle fragment to separate them. Two symbols could have two very high scoring cycle fragments that are separated by a few symbols where there is no cycling. Then I just add up the scores for all of the two symbol relationships to get the overall cycle score.
I also tabulate the fragment counts when comparing two messages with each other, so that I can see if I am making a message that has comparable fragment counts:
I have made a lot of messages, compared, and found that the 340 has about 25% to 30% random symbol selection. Meaning that when encoding, Zodiac stayed with his cycles most of the time, but chose a random symbol within any particular group of symbols that all map to the same plaintext about 25% to 30% of the time.
The first half of the 340 is somewhat more cyclic than the second half.
Yo smokie ….. KEEP AT IT…. Keep posting . I know of the cha
nging of the guard.. You can keep this thread going. Cheers
Smokie, thanks for the details on that calculation. I like your tabulation of cycle distributions. In my cipher measurements, I’ve been directly comparing various distributions by computing Euclidian distances between them.
For example, to compare n-gram distributions, I tabulate the counts in descending order. Cipher A has n-gram counts 5, 4, 2, 1. Cipher B has n-gram counts 4, 3, 1, 1. Then if we treat the counts as vectors, we can compute the distance as:
sqrt( (5-4)^2 + (4-3)^2 + (2-1)^2 + (1-1)^2) = sqrt(3)
I think this sort of distance calculation might be very useful when applied to your cycle tabulations. I will work something up and post my results soon.
