CIPHER STRUCTURE

[Jarlve]

I posted the new zzgrams file earlier, maybe you missed it.

Download zzgrams.zip

All ngrams in the zzgrams file have these:

ABCDE

ZBCDE
AZCDE
ABZDE
ABCZE
ABCDZ

ZZCDE
AZZDE
ABZZE
ABCZZ

Duplicates have been removed afterwards so it may seem some permutations are missing. Averaging won't be accurate either. If you can't get it to work this way then I think you should consider removing the wildcards as I did and adjust the ngrams to compensate. As I posted earlier this gave me a couple of amazing solves given the circumstances.

[smokie treats]

Jarlve,

I did use the zz ngrams on the last experiment. But I didn't see where there were any with letters between two z's.

For AAAH I have:

AAAH : 6 0
ZAAH : 6 0
AZAH : 6 0
AAZH : 6 0
AAAZ : 6 0
ZZAH : 6 0
AZZH : 6 0
AAZZ : 6 0

But what about ZAZH, ZAAZ and AZAZ?

When you have time. I'm just excited about the most recent experiment. I used a wildcard to represent a bunch of different letters and got a decent solve!

Smokie

[Soze]

I hope it works out for the two of you. I have been sitting back just watching you work. Haven't the foggiest idea what you two are talking about but, hell, I'm excited too.

Soze

[Jarlve]

@Soze, we are trying to debunk/prove smokie's wildcard hypothesis.

Jarlve,

I did use the zz ngrams on the last experiment. But I didn't see where there were any with letters between two z's.

For AAAH I have:

AAAH : 6 0
ZAAH : 6 0
AZAH : 6 0
AAZH : 6 0
AAAZ : 6 0
ZZAH : 6 0
AZZH : 6 0
AAZZ : 6 0

But what about ZAZH, ZAAZ and AZAZ?

When you have time. I'm just excited about the most recent experiment. I used a wildcard to represent a bunch of different letters and got a decent solve!

Smokie

Then we'll have so many duplicates which we don't know how to score! Also when a piece of text is scored with the ngrams it goes like this.

For 4-grams:

ilikeagoodmystery
ilikeagoodmystery
ilikeagoodmystery
ilikeagoodmystery
ilikeagoodmystery
ilikeagoodmystery
ilikeagoodmystery
ilikeagoodmystery
ilikeagoodmystery
etc...

But you may be right. Maybe I should recompile ngrams from a corpus after adding the same (upscaled) frequencies of "Z", "ZZ", "ZZZ".

[smokie treats]

Thanks Jarlve.

I was thinking about how Decrypto must compare short pieces of the message to the list of ngrams, and I understand what you are saying about duplicates and scoring issues.

I like the idea of upscaled z,zz and zzz ngrams in the future but don't want to waste your efforts in the present. Let me look at why Decrypto preferred double SS instead of double ZZ in the last experiment. I can look at the scoring of those ngraphs to see how they compare and report back to you about that.

In the mean time, we can think about how to score the consolidated z-ngrams. The trick for using z-ngrams to find the plaintext that a polyalphabetic ciphertext represents is going to have limitations. I will think about some simple experiments to explore those limitations. If you have any ideas for experiments, let me know and I will try them.

I have a busy day today (hopefully) away from the computer. Will be back soon.

Smokie

[smokie treats]

Addendum Analysis Experiment 4 ZZ Wildcard n-graphs

I checked the first three instances of wildcard 63 63 for score.

6363.png

Ngraphs for SS scored higher than the correct ngraphs and the zz ngraphs. Substituting 63 63 for the original symbols caused Decrypto to find a solution where we can see what the plaintext should be, but Decrypto DID NOT use the zz ngraphs when it solved.

So it seems that using the one z ngrams worked in a direct sense, but the zz ngrams worked indirectly because it made Decrypto look for ngrams with double letters that would work for a solution.

Jarlve, I agree that we would not know how to score ngrams with multiple wildcard symbols.

At this point I am going to have to re-group and think about the problem for a while. Like I said, I get sort of obsessed with the 340 for about a week or so each year. But I am going to have to rest for a while.

1. Zodiac used cycles but so far there is no way to determine which of the thousands of cycles are Zodiac made or random;
2. The +, q, B and F symbols (my 5, 19, 20 and 51) don't cycle with the other fifty nine symbols well and may have a special purpose;
3. These symbols are either 1:1 substitutes, fillers or polyalphabetic;
4. Even the most repeated perfect cycles have missing symbols, and it is possible that Zodiac used the +, q, B, and F symbols as wildcards for the cycle symbols (the Wildcard Hypothesis);
5. There is a way to use ZKDecrypto to analyze at least one symbol as polyalphabetic with modified ngraph files where the letter Z is substituted for each letter in each ngraph.

For mental health reasons, among others, I am going to try to stop doing this now, at least for a while.

Thank you very much for your assistance, Jarlve and doranchak.

Smokie

EDIT: My last idea was an application that people could download and use to help solve the 340. Merge alleged wildcards into one symbol, or do whatever else you think should be done. It would probably take years for one computer to check all of the possibilities. But with many computers working together, it could go faster. A master program that delegates and manages the tasks of all of the application programs, and application programs that work in concert with the master program and each other to solve. Sort of like how people can download an application to help SETI look at tiny pictures of the universe for radio waves from other intelligent life in the universe. Everybody who wants to participate could and it would truly be a group effort. I think that if a solution could be found that way, it would provide a great deal of satisfaction to all of the participants.

[Jarlve]

I agree on your summation and understand very well why you need to take a break. What you call obsession I call drive, and you literally drive yourself to an-almost-breaking-point where, you know from previous experience, you really need to take a break, like or dislike! It has happened to me probably close to a 100 times. The best way to recover from this is to disconnect from the cause.

The dude abides!

[smokie treats]

Well, unfortunately, I fell off of the wagon a little bit today. I still think that some of the symbols are polyalphabetic. I did some testing to try to figure out what symbols may be wildcards, some of the results I will withhold for now because they don't seem to show much and are not finished. But then I had an idea.

On June 3, 2015, Jarlve made some ngrams for me with one z. I made a coded message with a polyalphabetic symbol that ZKD1.2 solved but with Z's in place of the correct letters. See "Decrypto seemed to have corresponded the n-grams with the Z's in them and used them in the solution" at viewtopic.php?f=81&t=267&start=100.

So this time I just applied the 340. I lowered the value for E in INIT KEY from 8 to 7 and raised the value for Z in INIT KEY from 0 to 1.

Without the z n graphs, ZKD scored 32969 and the result was gibberish.

With the z ngraphs, the score was only 34249 in 15 minutes and the result was gibberish. But the program assigned symbol 19, or the "+" to letter Z:

340July8.png

340ZJuly8.png

O.k., so why did ZKD1.2 assign symbol 19, or the "+" to letter Z, but not assign any other symbol to letter Z? This was using the Z ngraph files, not the ZZ ngraph files.

I may do some more testing in the future, but is this a possible way to flush out the polyalphabetic symbol(s), if they are there? Is the + symbol the only polyalphabetic symbol? Are there other ways to test this hypothesis?

Smokie

EDIT: Are the scores for the z ngram files the only obstacle for finding a solution? If symbol 19, or the "+" symbol was the only polyalphabetic symbol, could this message be solved?

[daikon]

If symbol 19, or the "+" symbol was the only polyalphabetic symbol, could this message be solved?

I believe if you assign a new unique symbol for each occurrence of "+" in the ciphertext that would make it essentially a "wildcard". And I think that idea has already been tested and no solution was found. Either because "+" is not a wildcard, or because trying to solve a ciphertext of less than 340 characters long with 85 unique symbols is highly unlikely.

Ah, I remember where I saw it. Go to this page, and then in the "Available ciphers" list at the top select "Z340: pluses changed to unique symbols":
http://www.oranchak.com/zodiac/webtoy/stats.html

[glurk]

340.zodiac.uniplus.txt is that file. It has been included since the very first release of ZKD.

-glurk