Contest To Crack William Gibson Poem Agrippa

An anonymous reader writes "A new cracking contest to cryptanalyse a William Gibson poem. The electronic poem ('Agrippa') was written back in 1992 and self-encrypts after being displayed once. The person who successfully cracks the encryption will win a copy of every published Gibson book." The poem/program binary was recovered in 2008, but it looks like no one has managed (bothered?) to crack the code.

Comments

Well , I looked at it twice...

[Viol8]

... and it was still the same. Perhaps the 1992 tech doesn't work in my shiny new HTML5 browser? ;o)

Re:Well , I looked at it twice...

[SJHillman]

Crappy summary. The poem is not self-encrypting, rather a program displays the poem once and then encrypts it... it's that program that needs to be cracked. As far as I can tell, the poem itself is just a MacGuffin

Re:Well , I looked at it twice...

[Anonymous Coward]

The summary isn't clear. The posting on the web is just the text of the poem. Per the original linked-to summary:

While the text of William Gibson's elusive electronic poem AGRIPPA is widely posted around the Web, it has not been seen in its original incarnation — custom-built software designed to scroll the poem through a single play before encrypting each line with an RSA algorithm — since 1992.

Re:Well , I looked at it twice...

[polymeris]

Weird. From here I couldn't even see the decrypted version once. Just see undecyphrable verses.

And when they're done here

[jxander]

The next challenge is decrypting Finnegans Wake.

Re:And when they're done here

[Anonymous Coward]

I did that once.

I ended up with a Tale of Two Cities. Typo in the first sentence, "blurst" of times for some reason.

Re:And when they're done here

[Verdatum]

You stupid monkey!!

Re:And when they're done here

[Frosty Piss]

Everyone seems to assume it's a "master work" simpy because nobody can understand it.

Maybe it's just a bad book?

Losers will receive

[Anonymous Coward]

2 copies of every published Gibson book.

"win a copy of every published Gibson book"??

[KrazyDave]

There's your problem right there as to why no one has bothered. It's like that old joke - ".... second prize is two weeks in Cleveland, Ohio and the grand prize is one week in Cleveland, Ohio!"

Re:"win a copy of every published Gibson book"??

[realmolo]

It's true. I just re-read the "Sprawl Trilogy" in the last couple of weeks, just to see if "Count Zero" and "Mona Lisa Overdrive" were any better than I remember. They weren't. They're tedious-as-fuck.

"Neuromancer" is great, but Gibson went up his own ass after that.

Still, he wrote "Neuromancer", which gets him a lifetime pass.

I've cracked it

[mmarlett]

"Be sure to drink your Ovaltine."

Re:I've cracked it

[JustOK]

Be sure to salt your Crypto-routine

So in other words...

[TeknoHog]

They want you to hack the Gibson.

No need for new infrastructure

[vlm]

No need for new infrastructure, just post it as a project euler problem and when you "win" the blog/psuedowiki thing has a link to the torrent file for a collection of his books... I've uh, heard, that such a torrent exists.

Seriously though I've also heard second hand that if you really want to piss off Mr Gibson all you have to do it tell him you love his book iconic genre defining cyberpunk book "snowcrash". At least thats what I've heard. If you don't get the joke, if you ask people the name of a cyberpunk author then family feud style 90% of them will say Gibson but if you ask them their favorite cyberpunk novel a majority will say "snowcrash" which was actually written by Stephenson. I like Gibson's novels too, this is just funny how people assume the awesomest book must have been written by the awesomest author.

Not encrypted

[Dan East]

According to the diff of the disc image before and after the program runs (http://www.crackingagrippa.net/files/agrippa_diffs.txt) it's perfectly clear that the text is not being encrypted. The listing on the left is after the modification, and the listing on the right is the original disc image. A large portion of the disc (exactly 8,000 contiguous bytes) has been rewritten with only four different bytes: 0x41, 0x43, 0x47, 0x54.

Thus a very significant portion of the original information is lost during the "encryption". It sure looks to me like the program merely overwrites the poem portion of the data with one of four randomly selected bytes. The poem, as listed in HTML on the web page, is 9190 characters, which correlates pretty close with the amount of bytes being modified on the disc image.

Re:Not encrypted

[zamboni1138]

A large portion of the disc (exactly 8,000 contiguous bytes) has been rewritten with only four different bytes: 0x41, 0x43, 0x47, 0x54.

Aren't those hex for ASCII characters A, C, G and T? Isn't that the same four characters that are used in DNA sequences?

Re:Not encrypted

[WillAdams]

Correct.

The program was supposed to be over-writing itself w/ (randomly?) generated DNA information --- to match the DNA etchings / prints on the bundled cloth included w/ the physical original.

Re:Not encrypted

[Dan East]

Well in that case the only question is if that "DNA" contains some other message or meaning. Each of the 6,000 DNA "bytes" can be one of 4 values. Thus each "byte" can store 2 bits of information. That's 12,000 bits of data. Assuming uppercase letters only, it takes 5 bits per letter minimum to encode text (without any compression). That only allows for 2,400 characters, which is 1/4th of the poem text.

Looking at it another way, the poem contains 1649 words. That allows for 3.6 DNA "bytes" per word (6000 / 1649), which is 6 bits of data per word. I'll round it up to 7 bits per word to be generous. That's still only 128 unique values per word, which isn't enough to encode all the unique words in the poem.

I just don't see any way to encode a 9,000 word poem into 12,000 bits of data. If the "DNA" does have meaning then it would have to be an excerpt of the poem, or additional verses that aren't in the plaintext version.

The latter has my vote. If the author really is clever, then he came up with an algorithm that takes the original poem text and converts it into "DNA" looking data, which can further be decoded into text that contains additional readable text that completes the poem. If he pulled that off then he earns some major respect.

I found the problem

[Phoenixlol]

Perhaps "no one has managed (bothered?) to crack the code" because "The person who successfully cracks the encryption will win a copy of every published Gibson book." Just speculation, being unfamiliar with his work.

Re:What were security standards like in '92?

[djl4570]

DES is 56 bits and has been around since the seventies. Early browsers from c1995 used 64 bits because anything more required export licenses. That's what got Philip Zimmerman in trouble back in 1994 when PGP was first posted to boards and online services. Given that Gibson is a futurist he might well have used an early implementation IDEA which was first described in 1991 and is 128 bits.

Boring, though, isn't it?

[eyenot]

I mean, what's to stop a programmer, who isn't necessarily a heady cryptanalyst, from simply reverse-engineering the Mac application and figuring out exactly how it's done without looking at the poem itself (or the encrypted version) at all?

So, this isn't a cryptanalysis contest. It's a reverse-engineering contest. A cryptanalyst isn't given the actual encrypting mechanism, the original, and the cipher all out front and asked for an explanation. They just get the cipher and some reasonable expectation of what the original might possibly contain (the words "fuhrer" or "atom" for example).

So it's kind of boring for me -- a hobbyist with an ardent interest in cryptography -- to bother tackling the problem, when somebody with some familiarity with Macintosh machine language is going to have a severe advantage.

The contest caters to the wrong crowd and packages itself all wrong.