Between Cases

 

The detective was between cases. He had been for months. The police scanner, once normally buzzing with dire news and dangerous predicaments, was strangely silent. New cameras with excellent face and body recognition had been installed throughout the city. In truth there was a bit more to them than just analyzing looks. They could also make a profile of people based on their movement and intercepted various bits of electronic traffic. They could identify a person 98% of the time, even with a head to toe disguise and a feigned injury.

Crime had gotten so bad that people has ceased to worry about the privacy concerns and went to the other end of the spectrum. They registered their electronic devices and accepted the 24 hour surveillance. After everything was switched online, a flurry of arrests cut crime in half overnight. The next week cut it half again, and the once more the week after. Diminishing returns, but it was enough to prove the system had skill and teeth.

Outside the city, everything was still running as it was before. The criminals who were smart enough to leave were busy in the countryside. They worked honest jobs, quietly trying to leave the country before the system went over cold cases and archived footage and tracked them down.

There was a problem with misidentification in the system and while it started as a small hiccup, it seemed to be getting worse. Any of the tests run had it pass with flying colours, so it was hard to find the source of the problem. It seemed to be linked to a change in the weather, or lighting conditions, but it was hard to tell.

The police scanner was soon full of examples of sightings of known criminals, but none of the reports panning out. Suspecting he would be called in soon, the detective made a note of a few failed sightings and did some background research.

As expected, the detective was consulted after the techs couldn’t find the problem. The first thing that he noticed was that most of the effort had been directed at the system’s active components and not the passive database behind it.

The raw numbers from gait and facial markers meant little to most of the programmers, but the detective had read up on the values and could make basic sense of them. He scanned through the raw database looking for a clear example of what he suspected was off.

“There, Dan Nerish’s numbers are off. His left leg was damaged in a shootout, the walking style wouldn’t be the same for both legs, but this says they are. He also doesn’t have a mole on his forehead.”

The team did a check for Dan Nerish’s fake numbers and found that they were from a door to door salesman in the original database. Any time that man was seen, the system would flag him as Dan Nerish. The public’s appetite for privacy wasn’t totally gone, the AI wouldn’t record an image of a person, and only flag a criminal by name when passing on the sighting.

It was obviously an inside job, and the detective doubted that no-one had checked the database during the troubleshooting. The offending programmer was quickly tracked down and they confessed almost immediately. The question was why.

Like many other times, it simply boiled down to money. The system was good, too good to require many more updates, and the programmer in question had automated most of his job. It was easy enough to hire a low skill replacement for scanning in new people and essentially remove the biometric programmer from the equation. The system could have an unchallenged monopoly so future work prospects were looking thin as well.

One day a small restore error had duplicated the data for someone common under the name of an infamous criminal. The criminal in question had heard of the false sighting and botched police response and looked into the security company. The programmer had done a series of anonymous interviews and promised to shed some more light on the problem before the company squashed the story. It was easy enough to find the reporter who was going to break the case and bribe them to reveal the programmer. In turn bribing them into faking a few more false sightings.

As far as the programmer was concerned it was a win-win situation. He could mess up enough of the database to get a few more weeks ‘troubleshooting’ and get enough cash from criminals to make for a retirement package. He could eventually fix the problem but seed enough distrust in the system to make the public call for an overhaul. Or if management extended his contract, miraculously fix things with a small patch to the database. How quickly it was fixed and how unreliable it was could be balanced out on who paid for what.

He could bring out samples from the original, clean, database for the tests and keep suspicion off that part of the system. He didn’t count on anyone else being able to read the raw numbers.