—
A spin-off from your post: I suspect that plain texting is going to become the new Standard English–or, at least, American. Texters have zero difficulty (far from it!) in communicating with each other, and there’s no logical reason for future generations to be saddled with the arcane rules of "proper" English. So, text away, good people; and, to hell with the academics!
Hey moonrock,
AZdecrypt 1.0 now supports custom ngram alphabets, so you could try to explore it yourself if you like. I made some crude dutch, french, german, italian from small corpora and they work fine. It would be kickass if someone could create other high quality ngrams for AZdecrypt (languages and such).
Btw, I included some of your ciphers with AZdecrypt, do you mind?
Thanks
Hey moonrock,
I’ve pitted the 340 against a bunch of different languages in: viewtopic.php?f=81&t=3242
In a previous thread of mine, it was brought up that certain types of plaintext, such as spelled out numbers or lists of names, are "immune" to nested hill climbing because the plaintext deviates from standard English significantly, whereas nested hill climbing searches for increasingly English-like plaintext. This has also been discussed in the thread titled "Not all homophonic substitutions can be auto-solved." What if nested hill climbing were more broad, and that searching for standard language was only one form of nested hill climbing?
Names are solveable: viewtopic.php?f=81&t=3196&start=141
One of these alternative forms would be searching for numbers based on a small corpus of numbers, such as this: zero one two three four five six seven eight nine ten eleven twelve thirteen twenty thirty forty fifty sixty seventy eighty ninety hundred thousand million billion. This approach may not be able to decipher standard English plaintexts like the 408 cipher, but if, for example, the 340 cipher were just a list of numbers, then using nested hill climbing based exclusively on numbers would solve it instantly or near-instantly.
Spelled out numbers are also solveable, this is something we have looked at when daikon was still active.
It’s just a matter of creating ngram statistics that are suited for the task. It is somewhat of an endless street as you can come up with many specific ideas that make it hard for general purpose ngrams to solve.
There are good indications that the language information in the 340 has been spread out periodically. There is a bigram peak at period 19. Seemingly I was the first person on the internet that mentioned it (early 2015). It may previously have been overlooked because the period is quite high and that it cannot be inferred from the 340’s physical dimensions 17 by 20. Amazingly, there is a magic square in the Zodiac FBI files that when interpreted as a transposition matrix also produces a period 19 bigram peak.
Amazingly, there is a magic square in the Zodiac FBI files that when interpreted as a transposition matrix also produces a period 19 bigram peak.
So, I’m going down the rabbithole on this one: 340 = 18×18 + 4×4, so possibly 2 magic squares could be made with the characters in the 340. Has anyone generated the basic 18×18 square ("Siamese"method) and used it as a transposition matrix on the first 324 characters?
Amazingly, there is a magic square in the Zodiac FBI files that when interpreted as a transposition matrix also produces a period 19 bigram peak.
So, I’m going down the rabbithole on this one: 340 = 18×18 + 4×4, so possibly 2 magic squares could be made with the characters in the 340. Has anyone generated the basic 18×18 square ("Siamese"method) and used it as a transposition matrix on the first 324 characters?
I haven’t.
Here’s the magic square from the Zodiac FBI files. It doesn’t use the standard Siamese method because the horizontal component moves 2 squares at a time and it is specifically this which correlates with period 15, 19 when applied in a 17 by 20 matrix. Ofcourse then it wouldn’t be a magic square anymore, but it is particularly the reading rule that is of interest here.
