The standard CryptoQuip, as it appears in the comics, is a simple substitution cypher where one character of the solution is provided, and it's up to the player to figure out the rest from the provided cyphertext and the clue. It's a fun little puzzle to play, but that doesn't stop a more analytical approach.
This project is about a few different attacks on solving the general CryptoQuip puzzle in as short a time as possible. There are other versions of this same attack, and there are several different attack schemes. This just happens to be the fastest we've found.
I wanted to brush up on my ruby skills and posrting the Obj-C codebase to ruby seemed like a nice way to migrate the code to Swift, but brush up on ruby at the same time.
It all starts at Purdue University in the Fall of 1985, in the EE007 office of a few friends, and The Exponent, the daily campus paper, was publishing the CryptoQuip puzzle. Jim wanted to solve them himself, and I was simply not that interested, but I could see that there was a way to automate the finding of a solution if we simply had a set of reference words.
The first attempt at a solution was really a solution for a single word - which
was a simply grep command on the school's BSD unix machines hitting
/usr/dict/words and the pattern of the unknown word. That worked for
a while, and then I realized that all I needed was the file of words, and I could
write something on my PC with Turbo Pascal and I could do all the words
in the puzzle.
One thing lead to another... Jim wrote a shell script... I finished my Pascal version... and at this point, another officemate joined in: Bret. We each wanted to be able to solve the puzzle as quickly as possible - just by entering the puzzle and the given hint.
The current algorithm is based on the idea that each cyphertext word is really a simple pattern of characters. This pattern can then be used to filter out those words from the master list that simply don't match the pattern. At this point, each of the cyphertext words will have a series of possible words which then have to be compared to all the other possible words for all the other cyphertext words to see what words "match" and create a complete solution.
This is the purpose of the Legend. This is simply the potential key to decoding the puzzle. New Legends will be tried and see if they decode the plaintext words properly, and if so, we keep checking, if not, we discard that Legend, and try another.
There are a few other things - like what order to check the words, and how to prune the search tree, but those are all in the code as comments.
You can pull up the code in irb and then simply load the quip.rb file and
run the exercise function. It will run the solver on the test input and
return the result.
def exercise
w = get_words
ct = 'Fict O ncc bivteclnbklzn O lcpji ukl pt vzglcddp'
cc = 'b'
pc = 't'
dude = Quip.new(ct, cc, pc)
st = Time.now
puts 'starting attack...'
dude.attempt_word_block_attack(w)
et = Time.now - st
puts "solutions: #{dude.solutions} in #{et*1000} msec"
endWe didn't put in the effort to make this a RESTful endpoint with Sinatra, or anything like that... we just wanted to have a reference implementation that we could verify worked, and then get some metrics for the size of the codebase and the speed of execution.