from the I'm-not-familiar-with-the-sort-of-thing-you're-asking-for dept
Law enforcement agencies are generally pretty happy with their automatic license plate readers. It allows them to harvest millions of plate/location records without having to exit their vehicles, much less slow them down. It also allows them to spring from their cruisers with guns out and force non-car thieves into submissive positions while they perform the sort of due diligence that should have been completed long before the cops/guns exited their respective holders.
What they don't seem to like is anyone asking questions about the massive databases they're compiling or whether they've bothered to institute any minimization/privacy policies. When questioned, they usually talk about what a great tool it is for crime-fighting, even if said tool contains millions of useless photos entirely unrelated to criminal activity. Some even claim that every single photo in the database is integral to ongoing investigations and therefore cannot be subjected to minimization procedures, much less the pesky FOIA requests of surveilled citizens.
And sometimes, these agencies are so sure they like the tech that they can't even be bothered to determine whether it's actually doing anything to assist in the business of law enforcement. Stephen Gutowski at the Capitol City Project recently asked the Fairfax County, VA police about the effectiveness of its license plate photo database and got this 'FILE NOT FOUND' statement in response.
This letter is in response to your FOIA request in which you requested the number of ALPR records Fairfax County currently has on file. This number is constantly fluctuating, but as of 05/20/2014 at 1003 hours there were 2,731,429 reads in the system.
You further requested any available metric the county uses to determine the system's effectiveness. It was found that the Fairfax County Police Department does not possess any such responsive materials based on the information you requested.
The assumption here is that the system works. The Fairfax County PD occasionally posts arrests linked to ALPR database hits and… well, beyond that, the PD draws a blank. Presumably a handful of arrests justifies a multi-million image-and-location photo database. But this lack of self-assessment shouldn't be acceptable, not for an agency that has abused its technology in the past.
It came to light late last year that the Fairfax PD trolled political rallies to grab more plate data, racking up nearly 70,000 photos in five days. This abuse prompted a local lawmaker to push legislation aimed at severely limiting, if not completely eradicating, ALPR readers in his district. Not a bad idea, as far it goes.
Virginia law enforcement agencies aren't going to be happy with this move and they'll be able to mobilize a pretty powerful opposition. But these are the same entities that tried to bury info on plate readers back in 2009, simply because they felt the public might try to get the system shut down if they knew what was going on. But the lack of controls or any gauge of the system's effectiveness shouldn't be allowed to escape unnoticed, because the failure to monitor error rates and hits can result in catastrophic consequences for citizens whose plates trigger false hits -- something this system does at twice the rate of recoveries.
The license plate readers demonstrated a high error rate. Four ALPR vehicles used in Fairfax County over the course of five nights in February 2009 scanned 69,281 vehicles. The camera database produced twelve bogus hits and recovered four stolen vehicles, for a recovery rate of 0.6 percent and an error rate of 1.7 percent.
The technology can be used responsibly, but law enforcement agencies with tough minimization policies are almost nonexistent. And as we've seen twice in the last month alone, officers relying on faulty data aren't making an effort to verify database hits before attempting to effect arrests. Someone's going to be hurt or killed because of bad data, and hardly anyone in law enforcement seems to be concerned. If they did, strict policies on verification and disposal of non-hit data would be the rule, rather than the exception.