03-04-2025, 08:51 AM
(This post was last modified: 03-04-2025, 09:15 AM by Scarecrow.
Edit Reason: Rules
)
Sorry for late answers, been busy with many things lately.
First of all I have to say that by no means I understand everything the cipher is doing, I am not so qualified, but here is what I have understood or guessed how it is working so far. Attached is python script that should now handle most of the words at least. I may be wrong in many instances but something seems to work.
The cipher seems to mimic Voynichese as an idea, as it has many attributes of a conlang and uses various techniques like syllabic compression, null suffixes and bigram influence, maybe something else. Maybe it is constructed in this way to try to mimic some observed Voynich text properties, like the observed previous glyph influences the successive, last letter is common.
It uses also some techniques to defeat default analysis like dictionary attacks with word class substitutions, n-gram, frequency and entropy based analysis.
The ciphertext has two sequential PANGRAM twice plus something I do not know, fillers or something. There may be some errors as for example bar transforms to THE or THEN, making this not fully reversible. Maybe deliberate obfuscation or something I do not understand.
The decryption goes something like this (it does work for 80% of the cases so far):
1. Dictionary lookut: Some words are jsut direct mapped or I haven't found yet any way to decipher them, so it starts from that.
But if a word is not found the dictionary, the next step is to try remove specific suffixes from the end of the word, "nuy", then "yun", and then "ouy" in this order.
This suggests that these suffixes might have been added during encryption as some form of padding or indicator and are not part of the original plaintext.
2. Initial letter substitution: After removing the suffix,the first letter of the remaining word is the base or class of the word, look the letter in the word classification mapping. If it is found, it's replaced with the corresponding value. This is yet another tecnique, simple substitution, applied just to the first letter of many words.
The mapping I found so far was:
Cipher letter Plaintext Letter Class
F T Articles
q J Verbs
n B Adjectives
3. Last step is to find word inner parts substitution: The rest of the word (so after the first letter, sometimes excluding the last letter also), is then looked up in another dictionary. If the word (or a part of it that matches a key) is found, it's replaced with the corresponding value. This kind of phonetic substitution or abbreviation is found in most inner parts of the words.
4. Combine
Note:
The cipher uses bigram chaining, the last two letters of each word determine the next word’s transformations. For example: After a word ending in ig, the next word must start with q (e.g., maig → qar). Track sentence position (initial/mid/final) to handle positional variants (e.g., FaiT vs. bar = THE).
In short:
This decryption method can be described as a hybrid substitution cipher with rule-based transformations. It operates on ciphertext words using the following steps:
1. Direct Word Substitution: A predefined dictionary is checked for an exact match between the ciphertext word and a known plaintext word. If found, the plaintext word is used.
2. Rule-Based Decoding (for non-matches): If no exact match is found, the following rules are appl ied sequentially:
2.1. Suffix Stripping: Specific suffixes ("nuy", "yun", "ouy") are removed from the end of the ciphertext word.
2.2. Initial Letter Transformation: The first letter of the resulting word is substituted based on a fixed mapping of ciphertext letters to plaintext letters.
2.3. Body Decompression: The remaining part of the word (after the first letter) is then processed. Specific letter combinations within this part are looked up in a dictionary that maps these combinations to their likely plaintext expansions, often based on phonetic similarities or abbreviations.
3. Combination: The transformed initial letter and the decompressed body are then combined to form the final decrypted plaintext word.
Final observations:
Context Matters More Than Position, the cipher doesn't just care where a word is in a sentence - it cares what the text is talking about.
The encoding changes by subject which are at least, the pangram itself, properties (missing letters), typography uses and historical origins
For example, "THE" becomes different cipher words in each of these sections.
Vowel patterns seem to have some Rules - The middle parts of words use vowel patterns that compress information, more vowels in a row more compressed information
Different vowels compress different sounds.
The same pattern means different things depending on position (but deterministic)
Repeated vowels (like "iii") work differently than mixed vowels (like "iea")
The bigram chain effect is very predictable like in Voynichese.
The way one word affects the next follows consistent patterns:
Words ending with "iT" make the next word start with "K" (for "Q")
Words ending with "zz" make the next word start with "n" (for "B")
This creates a chain reaction throughout the text
As per new rules, all this with help of Claude and DeepSeek.
First of all I have to say that by no means I understand everything the cipher is doing, I am not so qualified, but here is what I have understood or guessed how it is working so far. Attached is python script that should now handle most of the words at least. I may be wrong in many instances but something seems to work.
The cipher seems to mimic Voynichese as an idea, as it has many attributes of a conlang and uses various techniques like syllabic compression, null suffixes and bigram influence, maybe something else. Maybe it is constructed in this way to try to mimic some observed Voynich text properties, like the observed previous glyph influences the successive, last letter is common.
It uses also some techniques to defeat default analysis like dictionary attacks with word class substitutions, n-gram, frequency and entropy based analysis.
The ciphertext has two sequential PANGRAM twice plus something I do not know, fillers or something. There may be some errors as for example bar transforms to THE or THEN, making this not fully reversible. Maybe deliberate obfuscation or something I do not understand.
The decryption goes something like this (it does work for 80% of the cases so far):
1. Dictionary lookut: Some words are jsut direct mapped or I haven't found yet any way to decipher them, so it starts from that.
But if a word is not found the dictionary, the next step is to try remove specific suffixes from the end of the word, "nuy", then "yun", and then "ouy" in this order.
This suggests that these suffixes might have been added during encryption as some form of padding or indicator and are not part of the original plaintext.
2. Initial letter substitution: After removing the suffix,the first letter of the remaining word is the base or class of the word, look the letter in the word classification mapping. If it is found, it's replaced with the corresponding value. This is yet another tecnique, simple substitution, applied just to the first letter of many words.
The mapping I found so far was:
Cipher letter Plaintext Letter Class
F T Articles
q J Verbs
n B Adjectives
3. Last step is to find word inner parts substitution: The rest of the word (so after the first letter, sometimes excluding the last letter also), is then looked up in another dictionary. If the word (or a part of it that matches a key) is found, it's replaced with the corresponding value. This kind of phonetic substitution or abbreviation is found in most inner parts of the words.
4. Combine
Note:
The cipher uses bigram chaining, the last two letters of each word determine the next word’s transformations. For example: After a word ending in ig, the next word must start with q (e.g., maig → qar). Track sentence position (initial/mid/final) to handle positional variants (e.g., FaiT vs. bar = THE).
In short:
This decryption method can be described as a hybrid substitution cipher with rule-based transformations. It operates on ciphertext words using the following steps:
1. Direct Word Substitution: A predefined dictionary is checked for an exact match between the ciphertext word and a known plaintext word. If found, the plaintext word is used.
2. Rule-Based Decoding (for non-matches): If no exact match is found, the following rules are appl ied sequentially:
2.1. Suffix Stripping: Specific suffixes ("nuy", "yun", "ouy") are removed from the end of the ciphertext word.
2.2. Initial Letter Transformation: The first letter of the resulting word is substituted based on a fixed mapping of ciphertext letters to plaintext letters.
2.3. Body Decompression: The remaining part of the word (after the first letter) is then processed. Specific letter combinations within this part are looked up in a dictionary that maps these combinations to their likely plaintext expansions, often based on phonetic similarities or abbreviations.
3. Combination: The transformed initial letter and the decompressed body are then combined to form the final decrypted plaintext word.
Final observations:
Context Matters More Than Position, the cipher doesn't just care where a word is in a sentence - it cares what the text is talking about.
The encoding changes by subject which are at least, the pangram itself, properties (missing letters), typography uses and historical origins
For example, "THE" becomes different cipher words in each of these sections.
Vowel patterns seem to have some Rules - The middle parts of words use vowel patterns that compress information, more vowels in a row more compressed information
Different vowels compress different sounds.
The same pattern means different things depending on position (but deterministic)
Repeated vowels (like "iii") work differently than mixed vowels (like "iea")
The bigram chain effect is very predictable like in Voynichese.
The way one word affects the next follows consistent patterns:
Words ending with "iT" make the next word start with "K" (for "Q")
Words ending with "zz" make the next word start with "n" (for "B")
This creates a chain reaction throughout the text
As per new rules, all this with help of Claude and DeepSeek.