kryptos beginner questions
Danica
I have a couple questions about kryptos, the first being how did
anyone know to use palimpsest and abscissa as keys?
thanks
Jim Gillogly
> I have a couple questions about kryptos, the first being how did
> anyone know to use palimpsest and abscissa as keys?
Speaking for myself, I didn't. I solved those parts of the
sculpture using classical methods after determining from the
statistics what kind of cipher it was likely to be. I recovered
the keys after finding the solutions. (That's the short form. :)
--
Jim Gillogly
Danica
> determining from the statistics what kind of cipher it was likely to be.
who solved it before you?
what are the "classical methods"?
how did you use stats to determine what kind of cipher it was ?(didn't
the other panel kind of give the type of cipher away?)
lastly, what does the morse code in the courtyard decipher to?
thanks for answering my questions, I think I may have just found my
new addiction :)
Jim Gillogly
>> I solved those parts of the sculpture using classical methods after
>> determining from the statistics what kind of cipher it was likely to be.
>
> who solved it before you?
In about 1992 (we think) a team of NSA employees worked on it
"evenings and weekends" over about a 9-month period. They solved
all but the last 97 characters, and didn't tell anybody outside
(including the CIA, evidently).
In 1998 David Stein, a CIA physicist, spent about a year of
"evenings and weekends" and solved the same bits as the NSA team.
He gave an internal seminar to CIA which was attended by (at least)
Ed Scheidt, the cryptographer who advised the sculptor. They
prepared to go public with ABC network, but the interviews they
had in the can were spiked. My conjecture is that an intel
screwup had just occurred (some nuclear tests by Pakistan and/or
India had happened without our having any advance information, if
memory serves), and the Company didn't want to give the impression
that one of their physicists was working on a puzzle rather than
these serious physics questions. The break stayed private.
I solved the same parts in one week of June 1999 and announced it
here in sci.crypt, and gave an interview to John Markoff of NYT,
who had (in my opinion) a good track record on technical issues.
As the NYT article came out, ABC thawed their footage of the CIA
break.
> what are the "classical methods"?
You'll enjoy reading David Kahn's "The Codebreakers", which will
give you a good grounding in methods used to break classical (i.e.
old) ciphers. In essence classical ciphers are those that can be
executed manually, or with simple mechanical equipment.
> how did you use stats to determine what kind of cipher it was ?(didn't
After the stats failed on the whole cipher, I started breaking it
up into smaller pieces and trying stats on each of them
individually. The start of the second section was pretty clearly
transposition, and one of the longer parts of the first section
between '?' symbols showed what could be a pretty good
polyalphabetic period, so I started trying in turn all the
polyalphabetics I knew, including Vigenere, Beaufort, Variant
Beaufort, Porta, Quagmire I, Quagmire II, and succeeded with
Quagmire III.
> the other panel kind of give the type of cipher away?) lastly, what does
No, it didn't give it away. Quagmire III should have the top
(plaintext) line keyed also. The tableau doesn't correspond to
the cipher. On the other hand, the tableau keyword (KRYPTOS) was
the correct one for the Quag III tableau. Of course the period
keys needed to be solved without that kind of clue.
If it'd been obvious, it wouldn't have taken ten years for the
first public break!
> the morse code in the courtyard decipher to?
Some of it is kind of palindromic, but so far as I know nobody's
seen anything in it other than its obvious face value meaning.
> thanks for answering my questions, I think I may have just found my new
> addiction :)
Let us know if you solve the last 97 characters!
--
Jim Gillogly
Danica
> seen anything in it other than its obvious face value meaning.
what is its face value meaning?
It seems like cryptography is now more about writing computer code
than just pencil and paper methods, is it just too hard to break most
codes by hand?
Danica
how they figured out the keys?
Jim Gillogly
> what is [the Morse code's] face value meaning?
See Elonka's website: http://elonka.com/kryptos/
>
> It seems like cryptography is now more about writing computer code
> than just pencil and paper methods, is it just too hard to break most
> codes by hand?
No - David Stein of CIA did his breaks by hand. I don't know
whether the NSA team used computers or not. I personally think
it's more fun to write programs to break things than to do them
by hand, but many of my attack programs are based on techniques
I used to execute manually.
Unless you mean modern cryptography: it's quite rare to see a
successful pencil-and-paper on a modern crypto algorithm -- the
recent attack on MD4 is the only exception that comes to mind.
--
Jim Gillogly
Danica
don't bother answering my question about what it translates to
Douglas A. Gwyn
> Unless you mean modern cryptography: it's quite rare to see a
> successful pencil-and-paper on a modern crypto algorithm --
Even if a manual solution is possible, there is no reason
not to use available tools such as computer programs.
This is especially true when it comes to making statistical
tabulations, sorting, cross-referencing, etc.
Jim Gillogly
> nice pics of the morse code by the way
> don't bother answering my question about what it translates to
at face value, so I'll point you more specifically to a page on
Elonka's website that lays it out:
http://elonka.com/kryptos/faq.html
See the question "Q: What do the morse code messages say?"
I haven't tried using them for anything having to do with the
decryption.
--
Jim Gillogly
Danica
Doing a bit of morse unravelling was fun, gave me some ideas...
I guess I'm just old fashioned about wanting to break stuff by hand.
How complex are the algorithms you used to break kryptos (how many lines of code)?
I assume C or C++ was used?
I should probably go do some reading about cryptography, thanks for the help guys!
Jim Gillogly
> I wasn't being cute about the morse code pics, I meant what I said at face value.
> Doing a bit of morse unravelling was fun, gave me some ideas...
> I guess I'm just old fashioned about wanting to break stuff by hand.
> How complex are the algorithms you used to break kryptos (how many lines of code)?
> I assume C or C++ was used?
Once one has broken it and knows what to look for, the algorithms
aren't horrendous. However, getting to where you can actually run
a program at it can be challenging. I don't have a breakdown on
how much of my code is directly Kryptos-relevant, because much of it
is part of a monolithic program that breaks (or attacks, anyway)
about 120 different kinds of ciphers. The C code for this whole
thing consists of:
headers: 500 lines
shotgun hillclimbing: 3400
cipher-specific hints: 700
cipher definitions: 6000
evaluation functions: 1000
trie control: 300
That comes to roughly 11,000 lines. The only two of those 120
cipher types that I actually used in the breakage were the Quag 3
and the double columnar transposition. Double transposition
wasn't really what Kryptos-3 was about, but it put enough together
that I was able to piece the rest together by hand. I didn't find
out what I think is the real way it was done (simple modular
counting) until a year or so later.
--
jim Gillogly
Douglas A. Gwyn
> ... Double transposition
> wasn't really what Kryptos-3 was about, but it put enough together
> that I was able to piece the rest together by hand. I didn't find
> out what I think is the real way it was done (simple modular
> counting) until a year or so later.
Indeed, it is common in cryptanalysis to detect some
causal pattern and utterly misinterpret it at first.
One has to have a certain amount of faith in the
process (derived from experience) in order to persist
when the initial assumptions turn out to not work.
And often the final solution isn't obtained with an
accurate model of the encryption process, but only
with a model that produces the same result. (That
phenomenon might obscure the actual key used, which
would be a pity if knowing the true key would provide
a clue about the key schedule or process used to
select memorizable keys.) Therefore, merely finding
the original plaintext is often not the end of the
analysis.
Douglas A. Gwyn
> Indeed, it is common in cryptanalysis to detect some
> causal pattern and utterly misinterpret it at first.
Speaking of which, is it well known that the 4th part
of Kryptos involves a period of 7? But how, I dunno
yet.
Jim Gillogly
It's certainly been noticed that there are more doubled
letters than expected (by a little), and, more importantly,
that all but one pair starts at the same point mod 7. My
first thought was that this suggested a period 14 poly with
the result taken out by a columnar transposition (with mostly
7-letter columns), and that a string appears at the same point
on two of the 14-letter rows leading to the doubled letters.
However, I haven't been able to make this work. There must be
*some* other explanation, or else I wasn't thorough enough in
my search.
--
Jim Gillogly
Danica
> > of Kryptos involves a period of 7? But how, I dunno
> > yet.
>
> It's certainly been noticed that there are more doubled
> letters than expected (by a little), and, more importantly,
> that all but one pair starts at the same point mod 7. My
> first thought was that this suggested a period 14 poly with
> the result taken out by a columnar transposition (with mostly
> 7-letter columns), and that a string appears at the same point
> on two of the 14-letter rows leading to the doubled letters.
> However, I haven't been able to make this work. There must be
> *some* other explanation, or else I wasn't thorough enough in
> my search.
Hey, I noticed the double letters too, does that rule out single
letter substitution?
Douglas A. Gwyn
> Hey, I noticed the double letters too, does that rule out single
> letter substitution?
No, but even such a short sample should have higher IC
if it is simple substitution.
Douglas A. Gwyn
> It's certainly been noticed that there are more doubled
> letters than expected (by a little), and, more importantly,
> that all but one pair starts at the same point mod 7.
More than that, period 7 creates columns with good chi IC
among too many pairs.
I toyed with the thought that it might be based on the
exhibited tableau with keyword KRYPTOS or WEBSTER, but
didn't get anywhere with a modest amount of experimentation.
Keep in mind that these messages were *meant* to be
solvable with manual methods, which limits the possibilities.
Jim Gillogly
> Keep in mind that these messages were *meant* to be
> solvable with manual methods, which limits the possibilities.
I've seen that said (e.g. by David Stein), but have seen nothing to
suggest that either Jim Sanborn (the sculptor) or Ed Scheidt (the
cryptologist) intended them to be solvable with manual methods.
Do you have an authoritative reference?
It's certainly the case that the first three parts are solvable with
manual methods.
--
Jim Gillogly
Mok-Kong Shen
Jim Gillogly wrote:
Just a wild guess: Could it be surely excluded that there
had been some errors? (One knows e.g. cases where buildings
were not constructed exactly according to their plans or
specifications and there had been thousands and more of wrong
proofs of FLT, some of which even done by math professors.)
M. K. Shen
Douglas A. Gwyn
> Just a wild guess: Could it be surely excluded that there
> had been some errors?
There were some minor errors in the solved parts,
but that didn't prevent solution. It would have
to be a gross error to make the puzzle insoluble.
One presumes that Scheidt decrypted the message
to ensure that it was correctly enciphered, in
which case any error would be in transcribing the
ciphertext into the sculpture. (E.g. reading a
P for a B, or dropping a letter.) I wouldn't be
too worried about the possibility.
Douglas A. Gwyn
> On Mon, 30 Aug 2004 01:50:53 -0400, Douglas A. Gwyn wrote:
>>Keep in mind that these messages were *meant* to be
>>solvable with manual methods, which limits the possibilities.
> I've seen that said (e.g. by David Stein), but have seen nothing to
> suggest that either Jim Sanborn (the sculptor) or Ed Scheidt (the
> cryptologist) intended them to be solvable with manual methods.
> Do you have an authoritative reference?
I think I heard it more directly, but all I could find
readily comes from the Washington Post article:
Sanborn and Scheidt wanted "to make something that could
eventually be deciphered or extracted, rather than
something that will never be done, ever," Scheidt says.
Also, Sanborn once said that he expects that eventually people
*will* read what he wrote, but that what he wrote is a mystery
in itself.
It doesn't seem likely that after three parts in very
classical (manual) systems, the fourth part would involve
anything exceedingly difficult, apart from its shortness.
Danica
> > It's certainly been noticed that there are more doubled
> > letters than expected (by a little), and, more importantly,
> > that all but one pair starts at the same point mod 7.
this seems like something that ought to be exploited
> More than that, period 7 creates columns with good chi IC
> among too many pairs.
what is "good chi"? How does one calculate IC?
> I toyed with the thought that it might be based on the
> exhibited tableau with keyword KRYPTOS or WEBSTER, but
> didn't get anywhere with a modest amount of experimentation.
why WEBSTER?
> Keep in mind that these messages were *meant* to be
> solvable with manual methods, which limits the possibilities.
personally, I firmly believe that this should be solvable manually
with enough ingenuity
Danica
Any ideas?
Mok-Kong Shen
Danica wrote:
[snip]
> personally, I firmly believe that this should be solvable manually
> with enough ingenuity
Whether manually or with the aid of a computer, 'enough
ingenuity' is the problem, I suppose. How much is 'enough'?
Maybe the author had assumed something that in his view
any 'ingenious enough' person would eventually think about,
but his assumption could turn out to be unreasonable, without
his being conscious of that. Note that the author, like
everybody, could be to some extent quite subjective in his
opinions.
M. K. Shen
Douglas A. Gwyn
> Maybe the author had assumed something that in his view
> any 'ingenious enough' person would eventually think about,
> but his assumption could turn out to be unreasonable, without
> his being conscious of that.
There is no reason to think that that happened.
Scheib devised the systems for Sanborn with the
intent that they be solvable, and the three that
were solved are in the vein of classic C/A
exercises. The only semi-plausible thing that
could render the fourth part uncrackable would
be that Sanborn applied it to a text that is
way too short to give the analyst a fair shot.
But we don't have any evidence that that happened,
either. My guess is that when it is finally
cracked, it will be by somebody versed in
classical C/A who happens to make a lucky guess
(about the key, the general system, or ???) that
enables further progress. That actually happens
a lot in cryptanalysis, and is part of the "art".
Mok-Kong Shen
One important assumption made in such cases is namely,
as you indicated, whether the text is long enough for
certain cues to be sufficiently plausibly discerned after
certain kinds of transforms that the author (again)
assumes that the analyst would eventually think of and
try. Such matters are all to more or less extent fairly
'subjective' from the very beginning, I am afraid, There
could be no way for the author to test for the
'objectivity' of his assumptions, unless the problem he
posed were in the class of the rather trivial ones. You
mentioned the lucky guess of the key. What if the author
assumed that the analyst would consider for that purpose
phrases from, among others, a certain well-known text of
the literature and that assumption turned out to be
overly presumptious? Anyway, I happen to know of cases
of examinations where the teachers apparently had certain
too high expectations about the capability of his
students with the result that some problems were solved
by nobody.
M. K. Shen
Danica
> of examinations where the teachers apparently had certain
> too high expectations about the capability of his
> students with the result that some problems were solved
> by nobody.
>
> M. K. Shen
Well, then the students aren't trying that hard or they aren't that
smart are they? If what you say is true then it would be a sad state
for the human race and our intellectual growth. I personally have so
many ideas about how to attack the fourth part, it will take me a very
long time to try them all. And I'm sure they are different from what
other people have tried. So hopefully we play this game together
until the einstein of kryptos comes along...
Mok-Kong Shen
Danica wrote:
>>Anyway, I happen to know of cases
>>of examinations where the teachers apparently had certain
>>too high expectations about the capability of his
>>students with the result that some problems were solved
>>by nobody.
> Well, then the students aren't trying that hard or they aren't that
> smart are they? If what you say is true then it would be a sad state
> for the human race and our intellectual growth. I personally have so
> many ideas about how to attack the fourth part, it will take me a very
> long time to try them all. And I'm sure they are different from what
> other people have tried. So hopefully we play this game together
> until the einstein of kryptos comes along...
Let me describe roughly a case of exam I happen to know, it
involved a rather sophisticated techique that is used in a
quite famous theorem (for the theorem carries the name of
the person who first proved it) in the field being examined
but was taught much much later in the course. Do you think
that it's o.k. for a teacher to expect that capability
from the students in an exam? BTW, while most people 'believe'
that Fermat didn't have a 'correct' proof of FLT, that's only
a belief. So you might believe in Fermat's claim and attempt
to prove FLT with techniques much more simple than its final
highly sophisticated proof done by Wiles, i.e. with elementary
number theoretical methods that were known to Fermat's time.
Good luck!
M. K. Shen