Within the last week, two things happened: I finally got around to checking out the Fox show Fringe, the first season of which I noticed sitting tantalizingly in the Playstation Store, and my Ars colleague Erica Sadun wrote an article exploring all the delightful little Easter eggs sprinkled throughout the show. In particular she devotes some special attention to the so-called “glyph code”—a series of weird images flanked by glowing dots that appear as interstitial bumpers before commercial breaks. If I hadn’t been way nerdy for crypto before I started writing about the NSA habitually, that certainly pushed me over the edge, and I couldn’t resist taking a stab myself.
Now, here’s the part that pains me just a bit: Erica did a whole bunch of work that ultimately enabled me to crack the thing in a couple of minutes, but stopped just a hair short of the solution. From where she left off, it’s actually incredibly simple once you make one crucial assumption. Alas, there’s no deep dark mystery about the show’s arc concealed here, and the solution’s actually a bit anticlimactic, but it’s below the fold for those who are interested.
So the code is nothing fancy: It’s a simple one-to-one, monoalphabetic substitution cypher. But it’s isolated words, not a sentence, so handy strings like “the” or “and” don’t recur. Crucially, there are a couple of letters missing from Erica’s transcription of the pilot episode glyphs and possibly an extra glyph for episode 3, at least as compared with the list here. Also, it looks like there was a flub in the glyphs aired for episode 5. Throw a couple errors into the mix and a dictionary attack on a string of characters with no breaks becomes computationally infeasible. (If you want a reasonably quick result from your laptop, anyway.) But it’s trivial if you know where the word breaks are.
Well, by “trivial” I mean “trivial when someone else has already written a really solid algorithm for brute-forcing a ciphertext with a probability-weighted wordlist, and you can just use their code.” The “someone” in this case is UC Irvine computer science prof David Eppstein, whose program and dictionary made short work of the string I gave it despite several errors in that ciphertext. Basically, Eppstein and Erica Sadun did all the real work here, and I combined the products of their efforts with a few minor tweaks. [Update: And it strikes me I’m being sloppy here; “brute force” implies blindly running through possible keys and checking against a dictionary, which is emphatically not how Eppstein’s very elegant program works—the curious should just read his explanation.]
Sometimes the answers that seem too obvious are correct: Each episode yields a single word, in most cases relating to a central theme of the episode. Glancing around the Internets, it looks like the key omission some other folks trying to solve it were making was a failure to take into account the orientation of the image and the position of the glowing dot that appears with each. If you don’t factor those in, of course, the cipher seems insoluble because you’re counting a whole series of distinct symbols as a single glyph. Anyway, the solutions for the episodes to date are:
3: AEGER [Latin for “sick”]
5: SURGG [should be SURGE?]
For those who want to play along at home, that makes the letter-to-glyph assignment as below. Not all letters have been assigned glyphs yet, obviously.
Between cracking that and writing this post, I’ve probably given this an hour, which is really more than enough, but if someone wants to Photoshop those together into a translation grid for easy reference so people can transcribe the word for the episode as they watch, I’ll happily post it here. (Done! See update below.)
Eyeballing the incomplete key so far, there do seem to be some obvious glyph clusters, and I should note the possibility that solving the substitution cipher is only step one. Think, for instance, of those acrostic puzzles you sometimes see in the New York Times Magazine, where you fill in responses to a series of clues, then rearrange the letters into an adjacent grid to form a quotation. For purposes of cryptanalysis we can just treat each glyph as though it were a distinct letter chosen at random. But the glyphs actually have internal structure too. They’re a combination of three features: Image, orientation (i.e. each image has a mirror version) and one of (at least) three positions for those glowing dots. Cracking the cipher puts those symbols in a natural order; conceivably there’s some further puzzle to be solved by analysis of that series. Or possibly not… again, I’ve probably given this one too much energy for now, but others might want to give it a look.
Update: Dennis at FringeTelevision has gone ahead and gotten his Photoshop on, producing a lovely, suitable-for-framing key to the glyph cipher. Apparently there’s a new episode tonight, and while I’ll have to wait for it to show up in the PS3 store, those of you with advanced “receiver” technology for your TVs can print out a copy and see what the magic word of the day is:
You can get a bigger version at the link above, or another printer-friendly key from the folks at Fringe Podcast.