Vitanuova for 2001 April 5

General

The mozilla not party (kind of like the mozilla dot party except, you know, not) was at Zeitgeist today. I had a great time! I saw a bunch of people I know and got to talk to Jack Moffit about Vorbis for a while.

Arms

My right arm got a bit sore again, but I got documentation about being discharged from treatment last fall, and we're sending that off to the insurance company.

Cipher

I'm trying without any success to break a encrypted message which is a very short message in English in some kind of cipher like a substitution cipher which can be worked by hand. But no luck. I think I have some evidence that it's not a simple substitution cipher: there is a word which has a pattern of letters like AABCCDE. It doesn't seem to me that there is any English word which has that pattern, so I think the cipher is more complicated than simple substitution.

How do people go about breaking ciphers given only the ciphertext? Bruce Schneier is always telling us that we should assume that the cryptosystem is known and only the key is unknown; here I don't actually happen to know the cryptosystem, but I still want to break it.

I suspect this might be a Vignere cipher -- I do know that it's a cipher that has a short English "password" or key. The Vignere cipher is one of the most obvious possibilities beyond simple substitution in that case. I know that the Vignere cipher is supposed to be readily broken, but I don't have a whole lot of ciphertext and I don't even member the standard technique for breaking it.

I did generate Vignere encryptions and decryptions of the ciphertext, plus slight variations, for every word in /usr/dict/words. So now I have this file of about 300,000 possible decryptions, and I don't know a good way to search for a possibly valid decryption.

Somewhere or other there is a research project with a good statistical test for the presence of English text. One approach is to figure out (somehow) the probability that a certain text could have been generated by a Markov chain model for English text. If the probability that the text could have been generated by the model is high, then perhaps the probability is also high that the text is English.

How did John Gilmore's DES cracker recognize that it had decrypted its text when it did?

Wind

It's incredibly windy here. (And cold.)