Connection between some observations
Posted: Tue May 02, 2017 2:35 pmI've examined the encyclopedia of observations (http://zodiackillerciphers.com/wiki/index.php?title=Encyclopedia_of_observations) once again and tried to find connections between the observations. Of course this had been done before lots of times but I wanted to do this on my own. Thereby I made an observation that may have not beed discussed yet. If someone had the same idea we can delete my thread of course. Since I don't want to post things that are already known I post only a summary and my conclusion. If you are interested I can post my sheets and detailed results later.
I've examined the following observations by enciphering Jarlves plaintext library and compared the results:
To put it in a nutshell:
I think (like most others do) that repeated symbols in columns and rows is an good evidence that z340 was substituted horizontally. A vertical substitution or pure gibberish is very unlikely (unless the plaintext was gibberish).
P15/P19 bigram peak is "higgs boson" significant like david posted before. The odd/even and upper/lower bigram bias is probably just a coincidence.
Now I had the following question:
Where is the connection between P15/P19 and the repeated symbols in columns and rows? If you untranspose z340 by P15/P19 you get very high bigram repeats but the difference between row and column symbol repeats vanishes. Either one of the two observations can be true at the same time.
To sort things out I made another experiment: I transposed a plaintext by P19 and substituted it afterwards. In the final cipher text you can still see a bigram peak at period 19 and you still have the high difference in column/row symbol repeats.
Conclusion
If my experiments are correct then I should have found an evidence that if z340 contains a transpositioning step this definitely had been done before the substitution. Otherwise P19 and row/colum symbol repeats would not behave like they do.
The cycle score should not be affected very much when substituting an already transposed plaintext.
What do you think?
By the way:
Since z340 and Jarlves plaintexts have 17 columns and 20 rows it is not very accurate to compare the symbol repeats directly. So I've counted the repeats in the rows and then transposed the ciphers into 20*17 before counting the symbol repeats in the columns.
I've examined the following observations by enciphering Jarlves plaintext library and compared the results:
- Repeated symbols by columns and rows
Odd/Even bigram bias
Upper/Lower bigram bias
P15/19 bigram peak
To put it in a nutshell:
I think (like most others do) that repeated symbols in columns and rows is an good evidence that z340 was substituted horizontally. A vertical substitution or pure gibberish is very unlikely (unless the plaintext was gibberish).
P15/P19 bigram peak is "higgs boson" significant like david posted before. The odd/even and upper/lower bigram bias is probably just a coincidence.
Now I had the following question:
Where is the connection between P15/P19 and the repeated symbols in columns and rows? If you untranspose z340 by P15/P19 you get very high bigram repeats but the difference between row and column symbol repeats vanishes. Either one of the two observations can be true at the same time.
To sort things out I made another experiment: I transposed a plaintext by P19 and substituted it afterwards. In the final cipher text you can still see a bigram peak at period 19 and you still have the high difference in column/row symbol repeats.
Conclusion
If my experiments are correct then I should have found an evidence that if z340 contains a transpositioning step this definitely had been done before the substitution. Otherwise P19 and row/colum symbol repeats would not behave like they do.
Largo wrote:Personally I don’t think that Zodiac transposed an already enciphered text since this would mean that he had to draw all those symbols twice. I don’t know…but I think this guy was lazy (that’s only an assumption). The crossed out „k“ in z340 confirms that he was not willing to write the whole cipher again after he made a mistake (I would have started again since I am pedantic). Maybe he decided to make sure the symbols are not as cyclic as in z408 because z408 was broken because of the cycled symbols (repeating bigrams like double L).smokie treats wrote:I agree and was also thinking the same thing. Encoding before transposition = more work. But, he only had the opportunity to do it once, and may not have thought that at first. I have always preferred transposition before encoding, but have never been able to reconcile the strange cycle score left right top bottom.
The cycle score should not be affected very much when substituting an already transposed plaintext.
What do you think?
By the way:
Since z340 and Jarlves plaintexts have 17 columns and 20 rows it is not very accurate to compare the symbol repeats directly. So I've counted the repeats in the rows and then transposed the ciphers into 20*17 before counting the symbol repeats in the columns.