"It was beautiful, kind of like abstract art"

July 16, 2008 11:08 AM

In March 2007, the FermiLab Office of Public Affairs in Batavia, IL "received a curious message in code" via USPS. In May 2008, scientists posted a facsimile image of the letter to their blog in the hopes of soliciting cryptologists to decipher the letter.

A partial solution began to appear by the very next day. Geoff Milburn (creator of the Homebrew Air Conditioner and employee of the Canadian Space Agency) noted that the first and last paragraphs appeared to be in base-3 and base-2, and got to work. By the 17th, he had experimented with possible mappings and reached a solution. Meanwhile, John Graham-Cumming whipped up a Perl program to experiment with the mappings and posted his Perl-decrypted version.

The text? "FRANK SHOEMAKER WOULD CALL THIS NOISE," and "EMPLOYEE NUMBER BASSE SIXTEEN."

Two months later, the central paragraph remains unsolved, and a Chicago Tribune article has rekindled the web's interest, and spread work and comments on the mystery through blogs and forums.


Too simple for you?

Well, there's work still to be done on the Zodiac Killer's November 8 1969 cryptogram, the CIA's Kryptos sculpture (previously), the Voynich Manuscript (previously) and the Shugborough House Shepherd's monument inscription.
And of course, there's always the Mayday Mystery (previously and previously).

posted by subbes (45 comments total)

Honest question: How would an expert cryptologist go about determining that a particular passage is actually written in a very difficult code, and is not merely a random assortment of characters?
posted by Faint of Butt at 11:15 AM on July 16, 2008 [3 favorites]

In 1610, thinking he had discovered two moons orbiting Saturn, Galileo composed a message:

ALTISSIMUM PLANETAM TERGEMINUM OBSERVAVI ("I have observed the most distant planet to have a triple form")

… and sent it to Kepler as an anagram:

SMAISMRMILMEPOETALEUMIBUNENUGTTAUIRAS

Remarkably, Kepler managed to "solve" this as a message about Mars, not Saturn:

SALVE UMBISTENEUM GEMINATUM MARTIA PROLES ("Hail, twin companionship, children of Mars")

The German astronomer had predicted that the Red Planet had two moons, and imagined that Galileo was confirming his belief.

[via Futility Closet]
posted by nasreddin at 11:17 AM on July 16, 2008 [23 favorites]

BE SURE TO DRINK YOUR OVALTINE
posted by optovox at 11:19 AM on July 16, 2008 [12 favorites]

also solve this tia
posted by East Manitoba Regional Junior Kabaddi Champion '94 at 11:19 AM on July 16, 2008

D-r-i-n-k m-o-r-e O-v-a-l-t-i-n-e...
posted by Dizzy at 11:19 AM on July 16, 2008 [2 favorites]

How would an expert cryptologist go about determining that a particular passage is actually written in a very difficult code, and is not merely a random assortment of characters?

One word modern jackanswer: Entropy
posted by DU at 11:24 AM on July 16, 2008

I'm both glad and mad this post exists. I love these but never have time to work on them.

Also interesting to drop in on UserFriendly after all these years. Still not funny, huh?
posted by DU at 11:25 AM on July 16, 2008 [1 favorite]

The first and third paragraphs look suspiciously similar to Woodstock talking to Snoopy. Was there more to Charles Shultz than we realized?
posted by Cranberry at 11:27 AM on July 16, 2008 [1 favorite]

Honest question: How would an expert cryptologist go about determining that a particular passage is actually written in a very difficult code, and is not merely a random assortment of characters?

I'm not an "expert cryptologist", by any means, but I took a class on it once. Usually they do it through analysis of character frequency, though some ciphers render this technique less useful than others.

on preview: that's what DU means by "entropy". Languages are very non-random in their use of characters, so much so that you can sometimes solve straight letter-substitutions using nothing but frequency analysis.
posted by vorfeed at 11:36 AM on July 16, 2008 [1 favorite]

Could one not identify the sender by checking all autopsy reports for anyone who died recently of prolonged chortling?
posted by CynicalKnight at 11:39 AM on July 16, 2008 [2 favorites]

Was there more to Charles Shultz than we realized?

Oh damn you. Now I have to go check or (more likely) remain haunted by the possibility for the rest of my life.
posted by aramaic at 11:40 AM on July 16, 2008

I notice all these geeks are treating it as a pure math/logic problem. Much talk of binary, hex, error-correction, etc. But I haven't seen a single question about the purpose of the message. Frank (and employee FC) apparently don't know anything, but how about this: Who would send a message like this to Fermilab? And why with the generic address? Did they not know anyone on the inside (then how did they know Frank?)? Or did they know the lab so well they knew exactly to whom it would be delivered? Maybe they have the wrong Frank, since he doesn't know anything about noise and the message references him like it should make sense.
posted by DU at 11:42 AM on July 16, 2008 [1 favorite]

ETAONRISH: a 'word' helpful both for distinguishing substitution ciphers from letter-shift ciphers and Wheel of Fortune.
posted by subbes at 11:44 AM on July 16, 2008

The text?

"I'M IN UR FERMILAB WASTING EVERYONE'S TIME"
posted by three blind mice at 11:45 AM on July 16, 2008 [5 favorites]

Obligatory link to Metafilter's own Robocop Is Bleeding's comment fable.
posted by Jofus at 11:55 AM on July 16, 2008

I know about character frequency; that would pretty much have to be the first thing to go in order to make a code really tough. What if I were to say that this:

gggggggggggggggggggggggg

is a code that could be deciphered given the proper key? Would you believe me and try to decipher it, or would you just kick me out of the lab?
posted by Faint of Butt at 11:58 AM on July 16, 2008

The Voynich manuscript is one of the coolest things ever. Obviously a bunch of gibberish rather than an actual code but still really cool.
posted by Justinian at 12:00 PM on July 16, 2008

Actually, by "entropy" I did not mean (solely) character frequency. I just meant, is there enough information (in the mathematical sense) in this sequence for there to be a message in there, but not so much that it looks like noise.

gggggggggggggggggggggggg

Does not pass that test. But that isn't to say it couldn't be a code. You could have used (something similar to) a one time pad and coincidentally added exactly the random numbers to your message such that every character ended up the same. In general, I think a OTP would render an entropy test (let alone a decoding) useless, but I'm already teetering on the edge of my knowledge.
posted by DU at 12:02 PM on July 16, 2008

Don't forget Codex Seraphinianus. (yes, nonsense, but such pretty nonsense)
posted by oonh at 12:02 PM on July 16, 2008 [1 favorite]

Also, they never found Herb, did they?
posted by ericbop at 12:26 PM on July 16, 2008

I saw Phillip Glass perform at Fermilab.
posted by tr33hggr at 12:40 PM on July 16, 2008

Entropy is in the eye of the beholder, though. The message:

gggggggggggggggggggggggg

might be entirely meaningful, once you understood that it is not a cipher, but a code: g24, that the codebook says is "WE ATTACK AT DAWN".

Similarly, although entropy analysis of the Voynich Manuscript suggests that it is too structured to be a cipher, there are other ways that it could be hiding information. For example, a grille overlaid on the text could reveal the actual message, while the rest of the text is just scribbled nonsense.
posted by SPrintF at 12:41 PM on July 16, 2008

I love these sorts of puzzles.
posted by dejah420 at 12:41 PM on July 16, 2008

Delta Green fans will of course realize that It’s a message from the Mi-Go, in disguise as Greys, and reads something like UR HEAD, PLZ, I NEEDS IT FOR MY SPACE BUCKET. Posting it on the interwebs pretty much blows the offer of a free trip to Yuggoth though.
posted by Artw at 12:44 PM on July 16, 2008 [1 favorite]

One time pads have been found to be perfectly unbreakable -- there is no information in the ciphertext that gives any clue to what is in the encoded text.

Being a special cryptography case means it's basically outside the realm of cryptography, and the only way to break it is via social instead of mathematical means.

A one time pad of

gggggggggggggggggggggggg

can be deciphered by the receiver and anyone who knows the platen, I guess, but again, that's not real cryptanalysis.
posted by chimaera at 12:57 PM on July 16, 2008

I saw Phillip Glass perform at Fermilab.
posted by tr33hggr at 12:40 PM on July 16

Maybe he dropped a page of his sheet music.
posted by newmoistness at 1:13 PM on July 16, 2008 [5 favorites]

Thanks, everyone. You've pretty much answered my question. If I'm understanding the principles properly, there needs to be a recognizable amount of signal, which can be quantified mathematically. If it can't, the sample is either gibberish or something (e.g. a code or a one-time pad) outside of the science of cryptology. Either way, there's nothing for the mathematicians to do about it.
posted by Faint of Butt at 1:33 PM on July 16, 2008

ALL THESE WORLDS ARE YOURS, EXCEPT EUROPA. ATTEMPT NO LANDINGS THERE. USE THEM TOGETHER. USE THEM IN PEACE.
posted by ZenMasterThis at 2:02 PM on July 16, 2008

How would an expert cryptologist go about determining that a particular passage is actually written in a very difficult code

Part of the fun is, you can't always tell--especially with a small sample of text. So the fun of it is you are left to puzzle out IF it is a code at all, if it is a code, WHAT type, and if you can make a guess about what type, what it actually says.

FWIW, in theory and if used with perfect discipline, one-time pad is indistinguishable from random numbers. But in practice, all sorts of mistakes can creep in--see for example the Venona project, which came about because the USSR re-used some of their one-time pads. (They're called ONE-TIME pads for a reason!)

So many times you are looking for those little clues or slip-ups that allow you to get a handle on what otherwise may be completely uncrackable.
posted by flug at 2:18 PM on July 16, 2008

I'm thinking the ink dots play a large role in this. They are all over the place, and look like noise. Hmmm
posted by geoff. at 2:51 PM on July 16, 2008

gggggggggggggggggggggggg

It's a simple substitution of "g" for "n", thus decodes to a 24-letter word meaning "constipation".
posted by Rumple at 3:19 PM on July 16, 2008 [2 favorites]

Well, the middle code looks hex. Whatever these strange symbols on top are
Just the last three symbols have not been assigned to hex.
Since he has used all hex symbols from 0-F except 1 and A the last three symbols are either

AFC
or
1FC

Well. let's perl a little bit...
F0BE58F2FD636C79D2E493e6afc
F0BE58F2FD636C79D2E493e61fc
posted by yoyo_nyc at 3:41 PM on July 16, 2008

Faint of Butt: I know about character frequency; that would pretty much have to be the first thing to go in order to make a code really tough. What if I were to say that this:

gggggggggggggggggggggggg

There are an infitude of ways this could encode a message. Generally, with "short" messages, it can be very difficult to determine whether there is any actual signal in the noise. And "signal" itself is quite a value judgement. But encoding a real message of telegram length that can't be cracked is surprisingly difficult without mechanical/computer aid.
posted by phrontist at 3:56 PM on July 16, 2008

the key to decrypting a message like this

gggggggggggggggggggggggggggg

is finding the fnords.
posted by empath at 4:48 PM on July 16, 2008 [1 favorite]

Markov chains are useful for analysis of a text where you have discrete letters, or sequences of letters. If you analyze the two- and three-character frequencies of a given sequence of text, and then sort by frequency, you can compare it to the frequency distributions of known languages, and have a chance of identifying the language in which it is written, identifying which (if any) characters refer to vowels and consonants, and a few other tidbits of information. The shorter the text sample though, the less likely this is to work out.

AFAIK the Voynich Manuscript resists this kind of analysis, indicating that it may be randomly generated using a table that evenly produces letter sequences, which led to the hoax theories.
posted by aeschenkarnos at 5:31 PM on July 16, 2008

Has anyone tried researching setec astronomy?
posted by elfgirl at 5:38 PM on July 16, 2008 [1 favorite]

In 1610, thinking he had discovered two moons orbiting Saturn, Galileo composed a message: ALTISSIMUM PLANETAM TERGEMINUM OBSERVAVI ("I have observed the most distant planet to have a triple form") … and sent it to Kepler as an anagram: SMAISMRMILMEPOETALEUMIBUNENUGTTAUIRAS

I can't believe I listened to the smart-ass who authoritatively informed me that this was Galileo's to-do list:

TEST ANAGRAM, MAIL IT; SIN, MUMBLE "E PUR SI MUOVE."
posted by dyoneo at 7:35 PM on July 16, 2008 [14 favorites]

BTW, the 'moons' Galileo discovered were the two sides ('ansae', Latin for 'handles') of Saturn's rings. With his primitive telescope, they looked like blobs. He also discovered the four big moons ('galilean satellites') of Jupiter, among many other things.
posted by lukemeister at 8:11 PM on July 16, 2008

Robert Hooke encoded his principle of elasticity:

This idea was first stated[1] by Robert Hooke in 1675 as a Latin anagram[2] "ceiiinosssttuv", whose solution he published in 1678 as "Ut tensio, sic vis" which means "As the extension, so the force."

This interesting page has more material on scientific codes.
posted by Tube at 11:31 PM on July 16, 2008

The shortest coded message I know is "One if by land, two if by sea." A single bit! Talk about your unbreakable one-time pads.

Of course, it wasn't the content that was being concealed, but the act of sending a message. The idea is that it's buried in all the other random lights that might be seen around Boston that night.

Similar techniques are used by spies for "dead drops" and other ways to pass information. A particular letter on a sign is turned upside down, for instance. Or Bob Woodward's (alleged) flower pot and morning newspaper signals to Deep Throat, as ways to call for meetings.

So you can see why sending gibberish can be useful for concealing the fact that you have a coded message buried somewhere. See also: numbers stations or steganography.

It's actually more useful, but of course a lot more work, to send false or junk content. You can use it to determine if your code has been cracked, whether your code network has been infiltrated/betrayed, or even to confuse the enemy over a longer period of time. This was done by the Allies quite a bit during WWII.
posted by dhartung at 11:32 PM on July 16, 2008

The better an encryption or data compression scheme is, the harder it is to distinguish from simple noise. Patterns are susceptible to analysis. Patterns are reducible.

Noise VS signal is the challenge.

(And that's the joke in the first deciphered paragraph, "FRANK SHOEMAKER WOULD CALL THIS NOISE". By not recognizing trinary, the signal looks like noise.)
posted by lothar at 8:11 AM on July 17, 2008

For those who cling to letter frequency, it should be noted that not all codes are straight substitution; the Enigma, for example, consisted of many, many subsequent substitution codes. Imagine using a different substitution scheme for each consecutive character in your message; it gets quite a bit harder to break. Or even recognize as anything but noise.
posted by kaibutsu at 8:57 AM on July 17, 2008

It's actually more useful, but of course a lot more work, to send false or junk content. You can use it to determine if your code has been cracked, whether your code network has been infiltrated/betrayed, or even to confuse the enemy over a longer period of time. This was done by the Allies quite a bit during WWII.

One of my favourite WW2 stories was when the allies took the corpse of some random victim of drowning, dressed him up as a naval officer from some ship that had recently been sunk (complete with false ID, letters from the wife etc), planted a cipher key in his pocket, then left him somewhere in the Mediterranean where the currents would ensure he'd be washed up and found by the Axis powers.

Then, they sent a bunch of fake coded messages using the cipher, leading the enemy to anticipate & prepare for a landing at entirely the wrong place & time. From memory, this little piece of trickery greatly helped the invasion of Sicily.
posted by UbuRoivas at 9:09 PM on July 17, 2008

Ubu - you're thinking of Operation Mincemeat, I think...
posted by Rumple at 9:26 PM on July 17, 2008

yes! unfortunately, i was totally wrong about the cipher key part.

it's a great story, anyway.
posted by UbuRoivas at 9:48 PM on July 17, 2008