The heart of the piece, though, is the encrypted text, scrambled, Sanborn says, by "a coding system that would unravel itself slowly over a period of time."
When he began the work, Sanborn knew very little about cryptography, so he reluctantly accepted the CIA's offer to work with Ed Scheidt, who had just retired as head of Langley's Cryptographic Center. Scheidt himself was serving two masters. "I was reminded of my need to preserve the agency's secrets," Scheidt says. "You know, don't tell him the current way of doing business. And don't create something that you cannot break—but at the same time, make it something that will last a while."
Scheidt schooled Sanborn in cryptographic techniques employed from the late 19th century until World War II, when field agents had to use pencil and paper to encode and decode their messages. (These days, of course, cryptography is all about rugged computer algorithms using long mathematical keys.) After experimenting with a range of techniques, including poly-alphabetic substitution, shifting matrices, and transposition, the two arrived at a form of old-school, artisanal cryptography that they felt would hold off code breakers long enough to generate some suspense. The solutions, however, were Sanborn's alone, and he did not share them with Scheidt. "I assumed the first three sections would be deciphered in a matter of weeks, perhaps months," Sanborn says. Scheidt figured the whole puzzle would be solved in less than seven years.
During the two years of construction, there were moments of intrigue and paranoia, in keeping with the subject matter and the client. "We had to play a little on the clandestine side," says Scheidt, who talks of unnamed observers outside armed with long-range cameras and high-intensity microphones. "We had people with ladders climbing up the walls of my studio trying to photograph inside," Sanborn says. He came to believe that factions within the CIA wanted to kill the project. There were unexplained obstacles. For instance, he says, "one day a big truckload of stone for the courtyard disappeared. Never found. I saw it in the evening, went back in the morning, and it had vanished. Nobody would tell me what happened to it."
Sanborn finished the sculpture in time for a November 1990 dedication. The agency released the enciphered text, and a frenzy erupted in the crypto world as some of the best—and wackiest—cryptanalytic talent set to work. But it took them more than seven years, not the few months Sanborn had expected, to crack sections K1, K2, and K3. The first code breaker, a CIA employee named David Stein, spent 400 hours working by hand on his own time. Stein, who described the emergence of the first passage as a religious experience, revealed his partial solution to a packed auditorium at Langley in February 1998. But not a word was leaked to the press. Sixteen months later, Jim Gillogly, an LA-area cryptanalyst used a Pentium II computer and some custom software to crack the same three sections. When news of Gillogly's success broke, the CIA publicized Stein's earlier crack.
http://www.wired.com/science/discoveries/magazine/17-05/ff_kryptos?currentPage=2