Slashdot Mirror
What Happens When Police License Plate Readers Make Mistakes? (theverge.com)
An anonymous reader writes:
The Verge reports that San Francisco Bay Area police "pulled over a California privacy advocate and held him at gunpoint after a database error caused a license plate reader to flag a car as stolen, a lawsuit alleges." Brian Hofer, the chairman of Oakland's Privacy Advisory Commission, was handcuffed and surrounded by multiple police cars, and says a police deputy injured his brother by throwing him to the ground. They were finally released -- 40 minutes later. But ironically, Hofer has been a staunch critic of license plate readers, "which he points out have led to wrongful detentions, invasions of privacy and potentially costly lawsuits." (California bus driver Denise Green was detained at gunpoint when her own car was incorrectly identified as stolen -- leading to a lawsuit which she eventually settled for nearly $500,000.) And at least one thief simply swapped license plates with an innocent driver.
The executive director of Northern California Regional Intelligence Center, a state government program, acknowledged that the accuracy rate of the license plate readers is about 90 percent, yet "added that in some cases, the technology has actually exonerated people, or given potential suspects alibis. But there is no way for the public to know just how effective the license plate reader technology is in capturing criminals" -- apparently because police departments aren't capturing that data. Only one of the region's police departments, in Piedmont, California, reported its "efficacy metrics" to the agency -- with 7,500 "hits" which over 11 months led to 28 arrests (and the recovery of 39 cars) after reading 21.3 million license plates. The license plate readers cost $20,000 per patrol car.
In Hofer's case, he was driving a rental car which had previously been reported as stolen but then later recovered -- though for some reason the police or rental car agency failed to update their database. But he criticizes the fact that "somebody could pull a gun on your because of an alert that a computer system gave them."
"They're just pulling guns and going cowboy on us," Hofer says. "It's a pretty terrifying position to be in....
"This is happening more frequently than it should be. They're not ensuring the accuracy of their data and people's lives are literally at risk."
The executive director of Northern California Regional Intelligence Center, a state government program, acknowledged that the accuracy rate of the license plate readers is about 90 percent, yet "added that in some cases, the technology has actually exonerated people, or given potential suspects alibis. But there is no way for the public to know just how effective the license plate reader technology is in capturing criminals" -- apparently because police departments aren't capturing that data. Only one of the region's police departments, in Piedmont, California, reported its "efficacy metrics" to the agency -- with 7,500 "hits" which over 11 months led to 28 arrests (and the recovery of 39 cars) after reading 21.3 million license plates. The license plate readers cost $20,000 per patrol car.
In Hofer's case, he was driving a rental car which had previously been reported as stolen but then later recovered -- though for some reason the police or rental car agency failed to update their database. But he criticizes the fact that "somebody could pull a gun on your because of an alert that a computer system gave them."
"They're just pulling guns and going cowboy on us," Hofer says. "It's a pretty terrifying position to be in....
"This is happening more frequently than it should be. They're not ensuring the accuracy of their data and people's lives are literally at risk."
*nt*
Rather than going in guns blazing and injuring people with excessive force, why not just pull the car over and talk to the people?
If they are going to be violent or belligerent it would be quite obvious.
The one time police pulled me over because I was driving the same kind of car as someone they were looking for, they just walked up, told me to keep my hands visible (this is sensible) and talked to me, calmly asked for my ID, ran it and said "you're free to go" once they realized I wasn't the person. No guns out, no "GET OUT OF THE CAR!!", no being wrestled to the ground.
Police have gotten way too gung-ho lately, it's time to dial that back a few dozen pegs.
Traffic court with limited rights for tickets / missed tolls if they miss tag you.
But ironically, Hofer has been a staunch critic of license plate readers
That's just wisdom, bearing itself out. Irony would be previous support/praise of plate readers on his part.
You can argue about the plate technology, but the obvious big issue here is that the police help unarmed suspects at gunpoint. We have a severe police hiring, training, and discipline problem.
Exactly. The police listened to a machine and responded like machines.
What's that all about? Every incident brings police from miles around. I watched a homeless person being interrogated by a cop at around 5 am. Within 15 minutes there were six patrol cars blocking other traffic while the cops stood around in a huddle. Some citizens gathered too, as usual.
These cops have nothing else to do? They know that the situation is under control yet they flock together at every opportunity. Are they lonely?
A smart entrepreneur would put together a mobile donut truck and a police scanner so he could be on the scene of every incident.
...omphaloskepsis often...
This is pretty much exactly the OPPOSITE of ironic.
He was already a critic of these devices - and now he has been provided with additional supporting evidence as to why they are bad.
It would’ve been ironic (in the colloquial sense) had he previously been a gung-ho supporter of police’s use of license plate scanners.
#DeleteChrome
Police are the one public sector job that should be banned from unionizing. They have guns. They don’t need a union.
Seems like the headline is misleading. The READER did not make a mistake. The DATABASE it was checking against was in error. This means that without the reader, if the plate number was queried the result would be the same.
This is important. The reader tech is not the culprit. The old fashioned database's accuracy is the problem. And this is the same problem today as it was 10+ years ago.
So, automated readers do not cause problems. They simply allow the police to concentrate on several tasks at a time rather than simply sitting and entering in license numbers. THIS IS A GOOD THING. This essentially provides more policing at a lower cost. AGAIN, THIS IS A GOOD THING.
You just know that someone is going to die because they get pulled over due to a false positive and some trigger happy brown shirt over reacts when the detained person sneezes and starts shooting?
Then the rest of the Police force will close ranks and defend it as a "justified" action and the worst the officer gets is some paid leave while it blows over while the victim's family gets nothing.
You know it is only a mater of time,
My car's license plate has an "O" and I once received a bill from the Bay Area FasTrak, where my car has never been. Fortunately, they included a picture showing that the offending plate has a "Q" and they have a way to send in an appeal -- via snailmail only.
In the USA, with more guns than people, of course people will often be armed. It's their right and everything.
No, law enforcement thinks it's at war with the population it says it's there to protect and serve. You can see that in many things, and this is one of them.
Anyhow, "90% accuracy" is both not enough information and really quite damningly low given that most by far plate readings ought to end up negative. Do the math. My conclusion is that those things aren't there for their stated purpose, but to have an easy excuse to play cop once more.
While this is technically legal, it would most certainly attract the attention of the officer behind the screen.
This is generally a Bad Idea, as it means they'll find a reason to pull you over, and make your life miserable from there.
[End Of Line]
What happens is we get a thousand minature versions of what happened to Tuttle/Buttle in the movie Brazil
Some drink at the fountain of knowledge. Others just gargle.
The $500K was an exception. Usually the victim doesn't get ANY recompense.
I think we've pushed this "anyone can grow up to be president" thing too far.
that a vocal opponent of license plate readers gets "accidentally" flagged as an ultra violent criminal? Small world, eh? What are the odds?
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
Actually, it went all the way to the Supreme Court that police in USA's job is not to protect the people. It's to prosecute crime.
https://en.wikipedia.org/wiki/Town_of_Castle_Rock_v._Gonzales/
They don't have to "find" a reason. You just gave them one. doing ANYTHING to obscure the plate, even with IR is illegal. period. full stop,
Well... I MIGHT buy the setup idea... Except it was a rental he was in.
That makes a setup a whole lot harder to pull unless you believer is like TV.
I'll bet my eye teeth that everyone in the Northern California Intelligence Center thinks that "90% accurate" means that there's a 90% chance that anyone pulled over by this system is driving a car they're looking for. In fact, it means no such thing. Not even close. You need three things to make that determination: the false negative rate of the test, the false positive rate of the test, and the probability that any random car you sample is stolen (the "base rate" of stolen cars you're looking for).
The truth is that you can pull someone over with a "90% accurate test" and there might easily be less than a 1% chance that it's someone you're looking for if the base rate is sufficiently small and the false positive rate anything but vanishingly small.
Let's suppose that the test is 90% accurate on positives, and 99.9% accurate on negatives, and that you're looking for 1 in 10,000 cars. You sample a stream of a hundred cars and you get these results:
True positives: 10 true positives in the stream x 90% accuracy = 9 true positives found.
False positives:99,990 true negatives in the stream x 0.1% = 100 false positives found.
Probability of a car you pulled over being one you're looking for: about 8.2%.
No law enforcement agency should be allowed to use a gizmo like this -- or a drug test for that matter -- unless the person interpreting the results passes a test in basic probability. The same should go for any expert witness in court.
Post may contain irony: discontinue use if experiencing mood swings, nausea or elevated blood pressure.
Not necessarily. Suppose that the database lookup gives erroneous results with probability (let's say) 10^-6 (for whatever reasons). Then if you are entering 1000 cars manually, you will have on average less than 1 false positive per year. If you harvest 10M license plates per day, you will be getting an average of 10 false positives *daily*. People are not going to be pleased by false accusations...
"The license plate readers cost $20,000 per patrol car." Pretty sure the total BOM for a device that will reliably do this 24/7 cannot be more than about $200, with maybe a week of developer time to string together some open source software to do this.
Plenty of privacy problems with ANPR, but thats not the problem here.
The problem is that American Police approach every situation with guns drawn. They are incapable of having a polite adult conversation with a person about an issue out of fear that person might have a gun.
Read the review and it really bad. The main person was in a car that had the plate correctly processed. The problem was the car had been reported stolen but after being returned was not cleared from the police database.
Then I looked at the source, the verge, and knew it was a worthless article missing up true events with false interpretation and events.
and say I'm dead, and the asshats that shot me get off scott free. I'll bet that covers 90% of the scenarios.
The militarization of the police was a huge mistake that should be corrected ASAP. Not holding my breath tho, the cops seem to be hiring a bunch of shoot first ask questions later cowards. See the guy in Florida from a day or two ago. 911 told him to flag down the police. He did, and the cop shot him through his fucking windshield. This my friends is cowardice in action with absolutely no repercussions, no matter how egregious.
At 10% failure rate it seems there should be a small niche business suing for wrongful stops. At $500K a pop, there is money to be made for victims, laywers, and plenty of motivation to make the system more accurate (another business?). The system would fix itself - capitalism at work, unless of course legislators start interfering by putting laws in place preventing the victims from suing, rather than fix the accuracy of the system (which is much harder than signing a new law).
Laws to handle this situation incoming in 3 ... 2 ... 1. Something along the lines of you not being able to sue the cops for computer errors. A more subtle approach would be for them to outsource the technology to someone else, if they haven't already, and deflecting the lawsuits to them. The required liability insurance is going to make for some very expensive technology.
Now, I've received a red light ticket where the ANPR mistook a 'Q' on a plate for an 'O' and mailed me the ticket with (evidently) no human intervention. Because the vehicle description that came up was for a late model red Chevy Suburban. And my vehicle is a 40 year old green Landcruiser.
Have gnu, will travel.
That's still no excuse to injure an occupant of the vehicle by pulling them out of the car and slamming them to the ground. Surely by that point it would be apparent if an occupant of the vehicle intended to shoot their way out of the situation. That and if you have them physically under control enough to slam them down, they already can't go for a gun, the actual slamming isn't necessary.
OTOH, pulling a gun on someone who knows they've done nothing to call for that is a GREAT way to make them dangerous by putting them in fear for their life.
Add to that the fact that the scanner is only 80% accurate in the first place and even if it reads correctly, the database may be wrong (as it was in this case) and you have a significant chance that the people you're interacting with are completely innocent.
The headline could have been, "What happens when police make mistakes?"
The answer is, lives are ruined, people die, and criminals get away. Now what happens when police are the bad guys? Maybe we should focus less on the technology and more on the fact that police will use technology against us. Inevitably.
There was a football player who asked this question once and lost his entire career over it. Our national habit of putting police on a pedestal has given us a militarized police force who believes it is above the law.
You are welcome on my lawn.
Cops kill more people than terrorists in USA
Yes, cops are professionals, terrorists are just volunteers.
Blank until
I guess it’s sexist, but “going cowgirl” means something completely different.
Blank until
What does it mean that the system is said to be 90% accurate?
If this is indeed a simple accuracy (sensitivity) measure, it means that 1 in 10 readings will be wrong. So, if someone is driving a stolen car, there's a 10% chance the system will incorrectly fail to identify him/her. That's the easy part :)
There are 250M cars in the US, and each year .7M are stolen. Let's assume an average stolen car is driven for a month. That would mean that about 1 in 5000 cars is actually stolen (which is probably a huge overestimate as I would guess most stolen cars are scrapped, exported, or abandoned pretty quickly).
Now, let's assume 50000 cars are scanned, of which 10 are actually stolen. This will give 9 true positives (90% x 10), and 4999 false positives (10%*49990). So the chance that a car is actually stolen would be 9/5008 or about a tenth of a percentage.
The more reasonable assumption is that a wrong scan will give a random different number plate, which most likely does not belong to a stolen car - the chance of getting falsely assigned a stolen car plate is 10% (error rate) * .0002 (prevalence) = .00002. So, the chance of getting flagged in a stolen car is 90% + .00002 = .90002. The chance of getting flagged in a non-stolen car is simply .00002.
Now, out of our 50k scanned cars, we still get about 9 TPs, and now get about 1 (.000002*49990) FP. In other words, it turns out 1 in 10 people are falsely accused (not a very good ratio for drawing guns if you ask me). Of course, this is assuming stolen cars drive the same amount of miles as non-stolen cars. If non-stolen cars drive 10x the miles (ie are 10x as likely to be scanned), because stolen cars are abandoned, scrapped, or exported, this changes to about 50/50.
(the obvious solution, of course, is to manually check plate and car make and color before drawing guns...)
This exact same situation could have occurred were a cop calling in the plate, considering the issue was in the database.
The question becomes, where does an expectation of privacy trump the goal of cops to protect society. And by protect society I mean to identify stolen property. Surely we'd prefer identifying stolen property by the plates vs "driving while black?"
Real story : I ran out of gas on a 2 lane highway. Cop comes up, gets license/registration. Then offers me a ride to the gas station. He - by rule - pats me down, and has me get into the back. I kiss his ass, ask for his card, and slime a ride back to my card.
The same types of rules would apply to the felony stop in TFA, only in that case, the best info the cops had was stolen car, which would implicitly mean that the passengers were a higher risk than I was. I would love to see cop cams of this stop, and come to an understanding of both sides.
One of my teenage kids went with a pack and were found "hanging out" in an empty house. The cop brought my kid to my house, and was let off with a warning. Which independent of my feelings of failing to raise my kid correctly, was the right cop response.
Fullest, disclosure, I *believe* that some double digit percent of cops - ten percent? - gravitate to the control and guns inherent with cops in the USA. We've talked of that re the SWOTing of the poor person who was shot on his porch. As I've said, that's my sense given my observations 1st and 2nd hand with cops.
But that's not a plate reader error. That sort of thing has been happening since cops carried hard copy lists of plate numbers to watch for, updated infrequently or typed by some fat-fingered slob.
Have gnu, will travel.
about police militarization.
One of the consequences of Getting "Tough on Crime" nobody talks about.
Hi! I make Firefox Plug-ins. Check 'em out @ https://addons.mozilla.org/en-US/firefox/addon/youtube-mp3-podcaster/
14. Police and sheriffâ(TM)s patrol officers
Fatal injuries: 14.6 per 100,000 workers
17. Taxi drivers and chauffeurs
Fatal injuries: 13.2 per 100,000 workers
That puts taxi drivers in roughly the same order of fatal injurries as sherrifs and patrol officers, in your country.
Do you see Taxi drivers (exerting their 2nd amendment and) pulling out tons of guns whenever their client seems slightly less calm ?
Then how come you expect patrol officers to pull out deadly equipment, not to defend their lives because they are under fire, but simply at the slightest *suspicion of potentially stolen car* ?!
Your country seems really weird...
"Sufficiently advanced satire is indistinguishable from reality." - [Tips: 1DrYakQDKCQ6y52z6QbnkxHXAocMZJE61o ]
the headline doesn't fit the story, there was nothing wrong with the police license plate reader and it didn't make a mistake, it worked perfectly. The mistake was that somebody forgot to update the database. None of the cases in the story support the notion the reader is at fault (so reading the actual licenseplate wrong), the problem lies with the database itself or criminals using duplicate licenseplates (well nobody can do anything about that until there is a way to make a licenseplate really unique so it can't be copied). An officer can't see if the person driving the vehicle is a person who stole the car or not, and having dealt with a lot of criminals who stole cars and didn't go quietly when stopped, you can ofcourse understand why cops react this way when stopping a carthief. Anybody can claim he/she is innocent.
In this case the people who forgot to inform the car was found should get a fine for not reporting it properly.
The article just underlines the fact the system actually works..
I'd really like to see the body cam video if Hofer's stop. The fact that he's an activist in Oakland, CA makes me wonder if he wasn't fully cooperative.
But in the end, the license plate reader did not make a "mistake", the car was indeed reported stolen.
"pro-cop, pro-military, pro-gun" You don't live in California, do you? a pro gun, pro military pro cop candidate couldn't get elected dog catcher.