But the vast majority of suppliers of these secure phones (one of the biggest is Crypto A.G. of Zurich) cooperate with law enforcement by selling their encryption schemes to both the National Security Agency and the British GCHQ (the Government Communications Headquarters, in Cheltenham, England, which is the British counterpart of the NSA). So even most encrypted phone conversations can be listened to by the NSA and GCHQ. International businessmen discussing illegal schemes and drug cartels discussing transactions all tend to speak carelessly over “secure” phones, not realizing that most of them really aren’t secure at all.
But this particular digitally encrypted format was unknown to either NSA or GCHQ. And that was the third peculiar discovery.
The scrambled signal was sent immediately to the Cryptanalytic Division at NSA’s Headquarters/Operations Building. There it was run through a Cray supercomputer, which tested the signal against all known encryption schemes. But the Cray came up blank. The signal wouldn’t break. Instead of voices speaking, there was only a bewildering sequence of ones and zeroes that the computer couldn’t comprehend.
This in itself was extraordinary. The NSA’s computers are programmed with the keys to virtually every known cipher ever invented, every mechanism to encipher that has ever been used. This includes any system ever used by anyone at any time in history, anything ever written about in a technical paper, in a book, even in a novel, any cipher that’s ever been even floated as a hypothesis.
As long as the computers are fed a large enough sample of the cipher, and the encryption scheme is known to the NSA, they will crack the code. Most digital signals are broken immediately. But after minutes, then hours of churning, the computers were stumped.
The NSA abhors the existence of any encryption scheme it doesn’t know. To a cryptanalyst, an “unbreakable” encryption is like an impenetrable safe to a master safecracker, an unpickable lock to a master lockpick. It is a challenge, a taunt, a red flag.
Two cryptanalysts-cryppies, as they’re called within the Fort Meade complex-hunched before a screen and watched with mingled fascination and frustration.
“Jeez, what’s wrong with this one?” George Frechette said to his officemate, Edwin Chu. “Everything’s processing except this one string. Now what?”
Edwin Chu adjusted his round horn-rimmed glasses and peered through them for several moments at the flashing numbers on the screen. “We got us a new one.”
“What do you say we have a look at it?” George suggested. “Play with it a little?”
“Sure,” Edwin said. “Hey, I’m there.”
CHAPTER FOURTEEN
Professor Bruce Gelman, a small, slender, balding man with a wispy beard, was an assistant professor of computer science at MIT with a national reputation in the field of electronic engineering. According to Ken Alton, he was also a legendary hacker skilled in the intricacies of telephony and one of the founders of the Thinking Machines Corporation.
He could have been in his thirties or forties; it was impossible to tell. Dressed in a woolen lumberjack shirt over a plaid flannel shirt, he did not look like a typical university professor, but then, computer types rarely did. His office was located in the Artificial Intelligence Laboratory in a tall, anonymous office building in Kendall Square, Cambridge.
“I thought you guys were up to speed on this stuff,” he said, sipping coffee from a giant plastic cup. “You’re telling me the FBI labs threw in the towel?”
“Basically, yes,” Sarah said.
Gelman rolled his eyes, scratched at his beard, and chuckled. “I see,” he said with exaggerated politeness, leaving no doubt what he thought of the FBI. “Of course, this technician you talked to is right: it’s not exactly easy to restore a tape that’s been erased. That’s true.”
She removed from her briefcase a black cassette sealed in a plastic evidence bag and marked with a number, took it out of the bag, and handed it to him.
He gulped some more coffee, set down the cup, and knitted his brow. “We could get lucky,” he said. “Might be an old answering machine. Or just a poorly constructed one.”
“Why would that help?”
“Maybe the tape wobbles up and down in the machine, relative to the heads. Possible the tape guides are loose, and the tape wandered up and down some.”
“That would make it easier?”
He shot his left hand out for the enormous cup of coffee, and accidentally tipped it over. “Oh, God. Yuck.” Pulling some sheets of pale-blue Kleenex from a plastic dispenser, he mopped up the muddy spill, which coursed over a stack of papers. “Yuck.”
He retrieved the enormous cup, managing to salvage half the coffee. “You see, that would leave us a stripe of recorded information above or below what’s been recorded over it.”
“And if the answering machine isn’t old, or the tape guides aren’t loose?”
“Well,” Gelman said, “tape is three-dimensional, right?” He slurped loudly from the coffee cup, then gingerly set it down. “It has a thickness to it. The front and the back surfaces of the tape are affected differently by the recording process.”
Sarah didn’t entirely understand what he was driving at, but nodded anyway.
“So you compare the front and back surfaces of the tape,” he went on, “to see if there are any traces of magnetic information on the back of the tape. Sometimes that works.”
“And if not?”
“Well, then there’s an effect called ‘print-through,’ where you find traces on one section of the tape of what’s been recorded on a section right next to it. So there are various places to look for data. I’m surprised your labs didn’t think of this.” He shook his head disapprovingly. “So we can scan the tape and reconstruct it two-dimensionally, using VCR technology.”
“Can you explain that?”
He frowned and looked down at the coffee-stained papers arrayed on the desk before him. “So, it’s like this,” Gelman said. “This is a technique I developed for a-another government agency, under contract. Oh, hell, it’s obviously the NSA. Anyways, normally an audio tape is magnetized, negative or positive, on a stripe, okay?”
Sarah nodded.
“But on a videotape, the information is laid down differently. It’s recorded on stripes put down at a transverse angle to the tape, in order to fit more information on the same length of tape.”
“Uh huh.”
“So when it comes time to play back, a VCR uses a helical-scan playback head to read that information. Meaning the tape head moves at sort of an angle across the tape, okay?”
“Okay.”
“So if you want to play back a really narrow stripe of leftover information that’s sort of on the edges of a broader band-the vertical information as well as the horizontal-you can use this VCR technology, a similar helical-scan playback device.”
He paused a moment, and Sarah nodded to encourage him to proceed.
“So the helical scan goes across the tape, transversely, moving up through the newly recorded stuff and then over the narrow band of leftover information-the stuff we’re interested in, right? So, at regular intervals, we have these little blips of the stuff we want. The rest is garbage.” Gelman spoke more and more rapidly, with growing enthusiasm. “So, then the question is, how do you sort the wheat from the chaff, you know what I mean? How do you separate out the sound you want from the sound you don’t? Well, what you do is, you write a program to differentiate it out, right?”