3.
Raid on the Pantry
Lily Dehn almost caught the anomaly, or infraction, or whatever it was. Actually, it was just a flicker of motion, right at the edge of her vision. In fact, she was never completely sure it happened at all.
She had been auditing the check-out/check-in logs of the company’s Documate® Inventory Control System for the annual Department of Defense review. And right when she wasn’t looking, a number on the screen mysteriously changed. Or she thought it changed. Actually, numbers on this screen were changing all the time, because she was inside the system that tracked every piece of software in the company’s inventory.
Dehn and the other programmers called this place “the Pantry,” because it was where you stored both the things you used all the time—like pepper, salt, and prepared soup mix—as well as the things you might use once and never touch again until it was time to throw them away—like that package of hojas de aguacate, dried avocado leaves, that she had bought three years ago to make Oaxacan mole for Cinco de Mayo.
The Pantry held everything, from major ongoing projects involving whole operating systems, components and modules, instruction set architectures, application program interfaces, test runs—even the drafts of manuals and marketing materials—down to odd fragments of code that one software engineer might write and slip to another when she finally threw up her hands and acknowledged that the problem she was working on was totally fubar, or “fouled up beyond all recognition.” The Pantry stored every piece of software and supporting data the company’s programmers had ever written, going back to the beginning of time and even beyond, to the pet projects of its founders from before Pinocchio, Inc., even became a registered company, and long before it went public and had acquired shareholders. The Pantry even contained a few classic, standalone doodles written in basic and fortran.
And that was the problem, because Lily Dehn was auditing that part of the inventory, down among the dead languages, archival stuff from twenty and thirty years ago: old, obsolete, inert, and of no interest to anybody. But that’s where one of the check-out numbers suddenly changed.
The error—what made Dehn think she was seeing things out of the corner of her eye—was that the number didn’t just go up, because a minute ago, while she watched, it had gone up, from 25 to 26. It was what you expected an activity counter to do. But then, a second later, that same number went down, from 26 to 25. Although it still might have been a hallucination, because the difference between the figures “6” and “5” was just a few pixels closing off the bottom loop. That, and the shape of the 6’s top curve, compared to the 5’s straight line. It might have been a flicker of the screen—except the monitor she was using had state-of-the-art thin-film transistors, and they simply did not flicker.
Worse, from Dehn’s viewpoint as archivist, was the fact that the supposed check-out action, both coming and going, carried neither a corresponding time/date stamp nor any signature authority. She clicked on the line again to verify this fact. According to all the protocols of the Documate® Inventory Control System, such a thing was supposed to be impossible.
Not that Lily Dehn actually cared about the event itself. If someone wanted to take a peek at an old piece of Sweetwater Lisp, which was the coding language for the project in question, that was their business—in Lily Dehn’s personal opinion. But the Department of Defense saw things differently.
Over the years, Pinocchio, Inc., had moved from industrial automata and artificial intelligence projects for the commercial market to more and more government contracting. The C-suite had found it more lucrative to invent and activate military drones, hardened robotics, mil-spec prosthetics, human capacity amplifiers, and tactical and strategic systems than fiddle around with household appliances and environmental control systems. All that commercial stuff was a business risk, because you had to satisfy a whole demographic of customers, and if one or two of your products failed in service, you had endless warranty issues, if not liability questions and lawsuits. Building robots and writing code for the military, on the other hand, was a piece of cake. You only had to satisfy the needs or whims of an isolated senator or full colonel, and now and then a review board. They took at most ten percent of your product output into testing and, if it passed, sent the rest into stockpiles and warehouses in places like Utah and Nevada. No one ever had to know what the actual failure rate in service was—unless they had a war, of course, and then everything was fubar anyway.
But the Department of Defense was bug-nuts about security, deeply afraid of the potential for espionage and sabotage. They wanted to count literally every line of code the company programmers wrote, whether the DOD had bought it or not, whether it was current or not. And so, about five years ago, Pinocchio, Inc., undertook a big project to install the Documate® ICS, clean out and burn down every other server drive, desktop and laptop hard drive, and thumb drive within reach, and put everything behind a firewall under seven levels of access control and with an automatic check-in and check-out log to count every time someone made an upgrade to, or took a copy of, a piece of software. And then every year some poor staffer—this year it was Lily Dehn—got the assignment to review the logs and write a report for the Inspector General’s office. Just time-serving, bullshit paperwork …
Until one of the check-out entries started counting in reverse.
And the anomalous entry left no log trace.
And that was bad!
Or maybe it was … just a tic, a hallucination, a trick of her optic nerve.
Well, they were paying her to do a thorough job. And Lily Dehn always tried to be honest in her work. So she probably should report it to her department head—for what that would be worth.
———
“Howard, do you have minute?” Dehn asked her boss.
Howard Spence was in charge of Programming Resources at Pinocchio, Inc. His department served as a talent pool for software designers and engineers who were temporarily between assignments to ongoing projects and contracts. It was not a place to linger for long in your career, else you got stuck doing odd jobs like minding the Pantry.
“Sure, Lily.” The man waved her to a seat, never taking his pouchy eyes off his computer monitor. It was angled toward his side of the desk, away from her. So Dehn had no way of knowing whether he was working on a piece of coding, reading his email, or playing Age of Empires. From the little grunts and grimaces he was giving off, she guessed it was the latter.
That was all the attention Lily Dehn had come to expect from this man. Or from most men. She knew she was never going to get by on just her looks. Blonde hair—but too short and unruly. Green eyes—but too wide and appraising. Full mouth with generous lips—but too likely to say the wrong thing. And her body wasn’t any prize, either—too skinny and lanky. She had long ago decided she was going to have to make her way on brains and hard work.
“What you got?” Spence asked vacantly, still not lifting his eyes.
“Something came up in my audit of the software inventory.”
“Oh, that …” he said, preoccupied. “You nearly done?”
“Yes, that—and no.” Dehn went on to explain what she had seen—or thought she saw—and why it seemed important. “High-security databases aren’t supposed to change themselves. And doing it without a log trace should be impossible.”
“Who’s working on the software that got charged out? You talk to them?”
“It’s an old project—I mean, like decades. I tried to track down the people involved. The team leader, Doctor Bathespeake, retired a dozen years ago and, given his age at the time, he’s probably dead by now. His two assistants, project engineer Jennifer Bromley and analyst/programmer Daniel Raskett, both have long since moved on to other companies. The only other name attached to the file was Steven Cocci’s, and his involvement was eyes-only.”
“The chairman of the board,” Spence said, and he finally glanced up at her. “And he’s dead, too.”
“Yes, sir.”
“So what is this piece of software that’s suddenly become so popular?”
“It’s not real clear from the metadata. The official title is ‘Multiple Entity Program, Series II,’ which is obscure enough. It’s written in Sweetwater Lisp, which dates it. That also suggests it was some kind of attempt at an artificial intelligence—back when people were still trying to achieve human-scale consciousness with simple coding structures, rather than combining neural-net architecture with dedicated hardware resources. The file size is miniscule, however, very small, very compact. … I have no idea what it does.”
“You didn’t download a copy and play with it?” Spence suggested.
“Oh, hell, no! I didn’t want to gum up the check-out logs.”
“Yeah, you’re right. Probably not a good idea.”
“So what do you want me to do?”
“About the anomaly?” he said. “Have you written it up?”
“Not yet. First, I wanted to talk to you, for … guidance.”
“So what you saw—you think you saw—is not official?”
“The record only shows me fishing around in the project metadata. But that’s well within my access for the auditing task. We could explain my snooping as a bored software engineer just exercising her curiosity.”
“Hmm.” Now his focus was on Dehn completely. She discovered that Howard Spence had very dark eyes, with almost no pupil, which seemed to beam x-rays right through her. “If we report this—what you think you saw, which is somebody hacking a secure database and stealing a piece of antique software, all without hitting a trip wire or leaving a trace—then the Pentagon will land on us like the Marines at Tarawa. They’ll probably call for a strip, recode, and reload of the Documate system. That’ll be no end of fun. And you, Miss Dehn, might find yourself with a permanent job in my department. Does that appeal to you?”
“Mister Spence, much as I value my time here …”
“Who has access to that Multiple Entity file,” he interrupted. “I mean, other than you, as the currently assigned archivist?”
“Well, the programming team would all have had ‘read’ and ‘write’ access, of course,” Dehn began.
Of the seven levels of control permissions, the lowest access was none, which meant the user could not even know that the software, associated documentation, or even their assigned folder in the inventory system even existed. Increasing access levels proceeded through browse, which meant the user could see the folder and metadata, but not read the stored contents. Users with higher permissions could successively read the content, annotate it, create a new version of it, and ultimately write over the original code and its documents. Anyone with assigned permission at one level could do any of the lower-order tasks as well. The highest level of all was delete, on the theory that the person who created the code in the first place could also destroy it.
“But only Doctor Bathespeake had ‘delete’ access,” Dehn concluded.
“So who now has read or write access?” Spence asked.
“Nobody. It’s a dead project,” she replied.
“And any ‘delete’ permissions?”
“Again, no one now living.”
“Well, there you’re wrong, Miss Dehn. You’re the archivist now. You have that level of permission over everything in our shop.”
“I … didn’t know that, sir.” She was shocked by the implications.
“It’s an awesome responsibility. You want to use it wisely.”
“Are you saying I should delete this Multiple Ent—”
“I’m not saying anything of the kind!” he cut in quickly. “That would be irresponsible of you. But then … it’s old code, isn’t it? No concern to anyone inside Pinocchio, Inc., anymore. No business of anyone outside the company, either. And a great deal of concern to the Pentagon if pieces of code start magically disappearing out of our inventory control system. But deleted code is gone, isn’t it? No possible target of a hack then, is it?”
“I think I see what you mean.”
“One of your duties as archivist is to maintain efficient use of the company’s server space,” Spence went on. “I’m hearing reports from the other software teams of fragmentation, even some isolated instances of misallocation. Obviously, we’re pushing some kind of safety margin—and the new server farm isn’t due to start up for another month. So you might want to clean up the Pantry just a bit … create a little more usable space.”
“I understand, sir.”
“Just a suggestion.”
Dehn only nodded.
Spence focused on his screen. “Nice chatting with you.”
“Likewise, sir.”
———
An hour later, Lily Dehn was staring at her own computer screen again. The folder titled Multiple Entity Program, Series II glowed brightly with the highlighting she had just placed on it. She had her fingertip poised over the delete key.
Dehn wasn’t fooled by that conversation with Howard Spence. He hadn’t given her any kind of order or authority to do this, just hints and suggestions. The long way around the problem of the unauthorized, unsigned check-out would be a formal inquiry. She would have to interview the IT department about the security of the company’s firewall. She would have to explore possible deficiencies in the Documate® Inventory Control System with the vendor’s technical support people and with the Pinocchio, Inc., team that had overseen its installation. She would have to leave a paper trail of requests and meetings for the Pentagon to discover and question—and then it would be “the Marines at Tarawa” anyway.
But if she pressed that delete key, Dehn knew she would be crossing a line—morally if not legally. The notion that the program was old and unused and of no concern to anyone anymore was just a guess on her part. She had never looked into it, so she couldn’t be sure. Maybe the Multiple Entity program was important. Even though the check-out logs showed no recent activity, MEPSII might still be the apple of somebody’s eye—after all, the late Steven Cocci had kept his eye on it. So, for the sake of thoroughness, at least, she ought to look into the folder herself. Besides, she was authorized, and after she deleted the folder, its metadata and its detailed log profile would go to Data Heaven along with it.
She moved her finger half an inch and clicked the return key to open the folder.
Inside, she found an orderly array of subfolders: “Program Notes,” “Source Code” in a couple of dozen numbered revisions, “Object Code” in as many revisions, and “Machine Language.” Curiously, the folder contained no platform or chipset variants for either the object code or machine language—which were the executable, live-action forms of the software, stripped of the source code’s symbology and with limited comment lines from the programmers. Usually, for thoroughness, a programming team produced different compilations designed to run on different system types. But MEPSII apparently came in only one flavor. That was odd.
She opened the Program Notes folder and browsed the documents inside. There she found a great deal of discussion, back and forth in archived emails, about how the MEPSII program was supposed to operate. But no one had bothered to make a formal statement or rationale—for the C-suite, for the Review Board, for the record—of what the project was actually supposed to do.
Dehn switched over to the highest-revved Source Code folder and examined its contents. She had only a passing familiarity with Lisp as a programming language. The textual representation of the code, as compared with its executable binary form, was in a series of human-readable statements of the linked lists, written out in alphanumeric characters and operational symbols. She had the time, so she might as well try to tease some sense out of the code statements.
The first thing she noticed was that the program was written in ten modules—numbered zero to nine, as any good programmer would count them—and that all the modules were apparently designed to stand alone. She could identify calls between them embedded in each module, but nothing held them together. She sensed a loose organization, like the backbone and ribs of a snake or a centipede.
The zero module seemed to be a stripped-down, truncated, least-moving-parts rendition of the whole structure. An image came to Dehn’s mind, from the old sci-fi movie Alien, where the monster’s mouth had teeth and a tongue, and the tongue had a mouth, and that mouth had an extensible tongue armed with tiny, silvery teeth. She shivered in spite of herself.
The number one module looked the most like any of the artificial intelligences Dehn had ever seen. It processed test statements and propositions and made calls all over the place. She couldn’t tell, just by looking, how intelligent the code actually was. But it was relatively small, written in an antique language, so her guess was “not very.”
The number two module appeared to have a collapsible data cache, referenced to the decision processes of the number one module and indexed to an internal command line with its own data cache labeled sampndx. Dehn had the impression of a snake’s stomach, but this one caught and held decision results and raw data instead of bugs and mice.
The number three module looked like a code dictionary, comparing binary structures and platform references that stretched far beyond the needs and expectations of a dedicated Lisp processor. Dehn could identify bits and pieces of various programming languages current at the time the MEPSII project was in development. Not all of them—in fact, almost none of them—were used in an artificial intelligence context. Some of them she recognized as bits of commercial security software and some as parts of operating systems.
The number four module included a powerful random-number generator, working off high-digit polynomials and attached to four separate interpreters. To Dehn, the code structure smacked of a complicated statistical evaluator—or maybe a picklock designed to outthink security systems based on public encryption keys.
She stopped examining individual modules at that point and just glanced through the remainder—until she came to the number nine module, the last one. This module was a sophisticated phage, a code wiper. But curiously, it had specific recognition and deletion steps for the binary code in the other nine modules and a recursion that finally attacked and deleted the phage module itself. This MEPSII software was designed to do its work, then disappear completely.
Dehn shook her head. She still couldn’t figure out what the program was designed to do, but she didn’t sense anything good. It looked like a virus, a piece of coding designed to infiltrate any one of a number of particular operating systems, written in various languages, and take over. But it was way too big and way too complicated to be just a virus. And if it was designed to be a computer virus or tapeworm, it would be an exceptionally ineffective one, easy to spot and easier to kill.
No wonder MEPSII had been left to rust in the archive. It was a dud.
Dehn backed out of the folders until Multiple Entity Program, Series II was once again a glowing line on her screen. She moved her finger back that half-inch. She could simply press the delete key, exert an ounce or two of pressure, and make this problematic piece of software and the mystery of its unauthorized check-out go away. Her action would of course leave a final time/date stamp in the log, along with her own signature authority for this extraordinary breach of archive protocols. If anyone ever noticed the missing files—or one day someone called for them—she would have a lot of explaining to do. And she wasn’t sure the expedient of opening some server space for ongoing projects would wash in any kind of formal investigation.
But with luck, the loss of the Multiple Entity software wouldn’t be noticed for weeks, or even years. By then she would be in another job, maybe even with another company.
Lily Dehn sucked in a tiny breath between her teeth.
She pushed the key, and the glowing line blinked out.