[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
VMs: RE: VMS Lookalike encipherment
> Just in case this hasn't been posted before here is a link
> to my latest work on a VMS type encipherment. I hope it
> is self explanatory.
If I recall correctly from 'The Codebreakers', many of the strategies you
describe were used in the most complex codes developed by the Argenti
family. In addition, they also used nulls and nomenclators. They also
eliminated doubled letters in the plaintext, because those were common
enough in vernacular Italian to create a weakness in the code, and they
usually assigned multiple allographs to the vowels to mask their frequency.
This is a small difference from the cipher table on your web page, where
you've assigned three allographs to s,t,u and v.
The Argentis made certain that their ciphers were reversible, too. So, for
example, if AB represented a certain letter, then A could not be a letter on
its own, but you might find AC or AD or AEA. They had a trick for coming up
with appropriate bi- and triliteral combinations that would have these
properties, but I don't remember what it was. I wonder if that wouldn't
create the 'ordered' effect though.
I once came up with a test for nulls, and I didn't identify any in the VM,
though I did identify them in files I generated that had nulls. The test
worked like this:
1. In a natural language transcribed alphabetically, the frequency
distribution of neighbors of any letter are going to be determined by the
phonological properties of the letter itself. So, for example, the
frequency distribution of letters that come before and after 'A' will be
different from the frequency distribution of letters that come before and
after 'T', and both will be different from the frequency distribution of
letters in the text as a whole. 'T' will only rarely be followed by 'M',
for example, while that is much more common for 'A'.
2. Nulls have no phonological properties. If they are randomly larded
throughout a text, especially in high frequency, then the frequency
distributions of the letters that appear before and after them should be
more or less similar to the overall frequency distribution of those letters
in the text as a whole.
I probably still have the C++ program I wrote to identify nulls somewhere.
As I said, I didn't see any in the VM using this test, but I did identify
them in files I generated.
A variation on this test could be used to identify different ways of
representing the same letter, if you could correctly identify which
uniliteral/biliteral/triliteral combinations represented individual letters.
If you randomly alternate A with 8, for example, the frequency distribution
of neighboring letters should be nearly the same for the two allographs.
To unsubscribe, send mail to majordomo@xxxxxxxxxxx with a body saying: