The Voynich Ninja

(18-10-2023, 09:14 AM)nablator Wrote: You are not allowed to view links. Register or Login to view.

(18-10-2023, 08:21 AM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.One note of interest, of all these labels "okal" is by far the most common in VMS, occurring over 500 times. If, hypothetically, we are talking about a codex inspired by Roman/Greek pre-Christian beliefs, Jupiter could be called Jove/Iove/Deus/Zeus (which could explain the angle of the chart on f67r2: this label is on top there, and putting Jove on top should be natural for Jupiter worshipers).

The labels for planets that we see in diagrams in Latin manuscripts often use the genitive case, that's why we get Veneris and Jovis instead of Venus and Jupiter. The modern conventions for writing Latin consonants v/j and vowels u/i differently did not apply of course: a cipher would use only u/i letters, not v/j.

Nobody ever called Jupiter/Zeus Deus, the Greek genitive would be You are not allowed to view links. Register or Login to view..

If there is a system for enciphering, the positional rigidity of Voynichese and low conditional entropy being what they are, it is unlikely that more than a few letters could be enciphered in such short labels, a mnemonic or abbreviation maybe?

I doubt there were any Jupiter worshipers as late as the 15th century. Classical antiquity was a huge source of inspiration for the Renaissance, but always in a Christian framework.

That's interesting. I have almost no knowledge of Latin or Greek. In any case, as far as I understand, Deus was an extremely common word for the God in medieval Europe, so if the Voynich manuscript is a medieval European creation in Latin, it could use Deus for God, even if the God in question is not the Christian god. What I have in mind is maybe some small local cult of Jupiter, inspired by Roman/Greek traditions, which I assume were known in the middle ages despite all the attempts to suppress alt religions. It's a stretch, I know, that's why I'm saying I'm not ready to posit (that is, defend) this hypothesis.

(18-10-2023, 09:14 AM)nablator Wrote: You are not allowed to view links. Register or Login to view.If there is a system for enciphering, the positional rigidity of Voynichese and low conditional entropy being what they are, it is unlikely that more than a few letters could be enciphered in such short labels, a mnemonic or abbreviation maybe?

When you say conditional entropy you probably refer to text ngram entropy (randomness of a character given n preceding or following letters). It doesn't apply directly to distance ciphers, because distance ciphers introduce the variable of, well, distance. So, "alar   al daiin" is different from "alar al   daiin" in a distance cipher. Having said that, okal has no obvious distance variations per se, however if it's a popular syllable combination, it could be that short and still encoded in full.

E.g., consider the cipher I designed to showcase how distance encoding can handle a large source alphabet with a small number of marks.

You are not allowed to view links. Register or Login to view.

It takes only 9 marks and three distances to encode one of 27 letters per 2.5 loci of the code. With the alphabet of the size of Voynichese it's possible to encode up to one of 70+ symbols per 2.5 loci or a single Latin letter per single letter of the script, keeping apparent text entropy low.

(edit: fixed the wording, as it sounded as if you can encode 27 letters per 2.5 loci and not one of 27 different letters)

[attachment=7781]

That may be, but it does not explain the frequency of the word, and in all sorts of places.
Hence certainly no name.
The more frequent a word, the more likely it is to be ambiguous. I know it from dialect, where one word covers several variants.
Example Latin - German. English translation.

(18-10-2023, 10:05 AM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.E.g., consider the cipher I designed to showcase how distance encoding can handle a large source alphabet with a small number of marks.

I need to study your cipher, but I'm too busy now.

In the meantime you can study my cipher (1 to many at enciphering, 1 to 2 at deciphering, can be disambiguated with some binary "distance" variable, several possibilities): search "antipolybius" or "zigzag path" on the forum.

(18-10-2023, 10:27 AM)nablator Wrote: You are not allowed to view links. Register or Login to view.In the meantime you can study my cipher (1 to many at enciphering, 2 to one at deciphering according to some binary "distance" variable, several possibilities): search "antipolybius" or "zigzag path" on the forum.

This one? You are not allowed to view links. Register or Login to view. 
I couldn't find the exact rules you used, is there any post or article where I can see them?

In any case, the first question I have for potential cipher candidates: are they relatively easy to read and write to warrant using them for a whole book. Is it possible to just read it off the page?

(18-10-2023, 10:20 AM)Aga Tentakulus Wrote: You are not allowed to view links. Register or Login to view.That may be, but it does not explain the frequency of the word, and in all sorts of places.
Hence certainly no name.
The more frequent a word, the more likely it is to be ambiguous. I know it from dialect, where one word covers several variants.
Example Latin - German. English translation.

Unless the cipher doesn't preserve the order of the syllables. A very simple experiment, taking the text of Opus Majus (I use it as the benchmark a lot) let's count how many times 'de' and 'us' appear within up to 4 characters from one another:


It's 1090 times. The whole text is:


words.

So, this combination appears once every 218 words. Not as frequent as okal in the Voynich manuscript, however if the Voynich manuscript specifically deals with Deus in some form, the number could be higher. Also, if this is a distance cipher at work, there could be other possible character combinations that when encoded would produce okal as part of a larger encoded string. I'm not arguing strongly in favor of this hypothesis, however I see no obvious counterpoint to it as of yet.

(18-10-2023, 10:53 AM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.This one? You are not allowed to view links. Register or Login to view. 
I couldn't find the exact rules you used, is there any post or article where I can see them?

Yes, there are a few other posts on the next page. There could be complications, special rules for apparent ligatures and gallows. In the experiments I did there were no special rules, but many null words. Now I realize that null words are not necessarily a bad thing.

Quote:In any case, the first question I have for potential cipher candidates: are they relatively easy to read and write to warrant using them for a whole book. Is it possible to just read it off the page?

Ease of use was the main design constraint. To cipher, it requires a good look at a table to find a zigzag path that goes through the letters of a chunk of plaintext, which would explain the frequent small horizontal and vertical offsets, as if the scribe had been interrupted and had to take a look elsewhere before continuing. Deciphering is done by following the zigzag path defined by Voynichese glyphs interpreted as 2D coordinates (xyxyxy...) on the table. Not as easy as a simple mono-alphabetical substitution cipher but not as difficult as using a large nomenclator key, or cipher wheels, or code book, because the letters and Voynichese glyphs are easy to find in the table with a little practice, a glance would be enough.

(18-10-2023, 12:07 PM)nablator Wrote: You are not allowed to view links. Register or Login to view.Ease of use was the main design constraint. To cipher, it requires a good look at a table to find a zigzag path that goes through the letters of a chunk of plaintext, which would explain the frequent small horizontal and vertical offsets, as if the scribe had been interrupted and had to take a look elsewhere before continuing. Deciphering is done by following the zigzag path defined by Voynichese glyphs interpreted as 2D coordinates (xyxyxy...) on the table. Not as easy as a simple mono-alphabetical substitution cipher but not as difficult as using a large nomenclator key, or cipher wheels, or code book, because the letters and Voynichese glyphs are easy to find in the table.

What would it take to break it? Let's assume, the plaintext is Latin (I suppose we can just try the top candidate languages one by one). As far as I understand the principle, there is a rectangular table of letters and we skip from a square to a square writing down how many squares we skip horizontally and vertically each time. If the table is kept the same for the whole book or at least for large portions of the book (and I think it's really pointless to consider other possibilities, as they gradually approach the one-time pad in being theoretically unbreakable), can we try to identify any patterns that would betray some of the properties of the plaintext? How do we pick where to start on the table? It would be impossible to decipher this without knowing the starting square, I suppose.

(18-10-2023, 12:21 PM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.As far as I understand the principle, there is a rectangular table of letters and we skip from a square to a square writing down how many squares we skip horizontally and vertically each time.

It would be a variant using relative coordinates (relative to the current position on the table). I used simple Cartesian coordinates, absolute. Space = 0, o = 1, e = 2,  etc. in no particular order, but I found that some orders worked a lot better than others - more efficient.

Quote:If the table is kept the same for the whole book or at least for large portions of the book (and I think it's really pointless to consider other possibilities, as they gradually approach the one-time pad in being theoretically unbreakable), can we try to identify any patterns that would betray some of the properties of the plaintext?

My cunning plan is to try and crack my own enciphered Latin texts. I'm not there yet because I'm not happy with the special rules (ligatures and gallows) that I plan to add.

Quote:How do we pick where to start on the table? It would be impossible to decipher this without knowing the starting square, I suppose.

The lines start and end at the (space, space) coordinates (corners of the table), see: You are not allowed to view links. Register or Login to view.
"Vords" (Voynichese words) start and end at a border of the table.

(18-10-2023, 01:07 PM)nablator Wrote: You are not allowed to view links. Register or Login to view.It would be a variant using relative coordinates (relative to the current position on the table). I used simple Cartesian coordinates, absolute. Space = 0, o = 1, e = 2,  etc. in no particular order, but I found that some orders worked a lot better than others - more efficient.

If the labels in the manuscript are indeed labels (each corresponding to some plaintext words or numbers), and sometimes they are located without any obvious sequence on the page (e.g., star labels), does this mean this method will have to restart at each label? What would be the correspondence between the encoded label length and the plaintext length? If I understand the method correctly, the encoded label is between N and 2 * N characters long?