FBI, ICE using state driver’s license photos for facial-recognition searches
washingtonpost.com
washingtonpost.com
> The FBI said its system is 86 percent accurate at finding the right person if a search is able to generate a list of 50 possible matches, according to the GAO. But the FBI has not tested its system’s accuracy under conditions that are closer to normal, such as when a facial search returns only a few possible matches.
What the GAO study[1] actually said:
> However, we found that the tests were limited because they did not include all possible candidate list sizes and did not specify how often incorrect matches were returned. ... The FBI’s detection rate requirement for face recognition searches at the time stated that when the person exists in the database, NGI-IPS shall return a match of this person at least 85 percent of the time. However, we found that the FBI only tested this requirement with a candidate list of 50 potential matches. In these tests, 86 percent of the time, a match to a person in the database was correctly returned. The FBI had not assessed accuracy when users requested a list of 2 to 49 matches.
According to FBI, a smaller list would likely lower the accuracy of the searches as the smaller list may not contain the likely match that would be present in the larger list.
In other words, their acceptance test procedure was gamed from the beginning.
Take something really simple about percentages. What does 0.1% mean? Only 1 in 4 people know this means "1 in 1000".
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3310025/
> Gigerenzer et al show how only 25% of the general population could correctly identify 1 in 1000 as being the equivalent of 0.1%.
Take something a little bit more complex, such as the relative increase in risk vs the actual total risk. We know that most people do not understand what a 75% increase in risk means in real terms. But most people do understand a simpler explanation: of people who don't eat $THING we'd expect to see 1 person out of 1000 people over ten years developing a disease, but if 1000 people all eat $THING every day over ten years we'd expect to see about 2 people developing the disease.
See also cancer screening: (This is a good useful link that rattles through most of what I'd want to say) https://www.nice.org.uk/guidance/ng34/evidence/expert-paper-...
“If you participate in breast screening, you will reduce your chances of dying from breast cancer in the next 10 years by 24%” versus
“If you participate in breast screening, you will reduce your chances of dying from breast cancer in the next 10 years from 37 in 10,000 to 28 in 10,000”
BMJ has more about relative vs absolute risk: https://bestpractice.bmj.com/info/toolkit/practise-ebm/under...
Here's a final example. For some time we knew members of the public couldn't do this, but we thought healthcare professionals could. Turns out that they couldn't do it either. Both groups find it much easier if you convert this into natural numbers and probability trees.
"A machine has been invented to scan a population for a disease. The machine is good but not perfect. If you have the disease there is a 90% chance it will return positive. If you do not have the disease there is a 1% chance it will return positive. About 1% of the population have the disease. Mr Smith is tested, and the test comes back positive. What's the chance Mr Smith actually has the disease?"
(This is from "Reckoning with Risk" by Gerd Gigerenzer).
Most people cannot get the right answer from this question.
If you reword the question they can.
"Think of a group of 100 people. 1 of them has a disease. The entire group is screened. The one person who has the disease tests positive. Of the 99 people who don't have the disease one person will also test positive. How many people of those who test positive have the disease?"
You can also show this as a probability tree. https://imgur.com/a/JWVQRxI
Two books I recommend are "Reckoning With Risk" and "Risk Savvy", both by Gigerenzer.
So much of forensic science is a sham, we claim as a country to uphold a system of justice whereby you're innocent unless proven guilty beyond a reasonable doubt, yet so many prisoners on death row have been exonerated, some posthumously even when there's no real incentive to look for evidence of innocence at that point.
Who wants to bet that prosecutors are going to start using flawed facial recognition results as if they are equivalent to a victim picking out a suspect from a lineup of 10 people? "There was a 99% chance of a match"
Maybe prior to the early 90's when the technology and chemistries were still kinda crude and not every lab could afford accreditation, but definitely not in the US in the past 10 years.
To be clear I am not arguing about the philosophy of if it's ethical to use DNA databases or facial recognition from driver's license databases. I'm saying comparing the use of DNA evidence to using facial recognition on a driver's license database doesn't make sense.
Oh you know its both
The accuracy from most methods is between 99.2% - 99.8%, but the problem is that the training samples are too easy and controlled. It's sensitive to lighting. Google's most recent paper [1] on Facenet found 99.63% on the easy Labeled Faces in the Wild (LFW) dataset, and an impressive 95.12% on the Youtube faces dataset, presumably a much more difficult dataset.
Whereas the DMV only has 1 picture of a suspect so it will result in more false positives.
We already have this. Look up the false positive rate of drug sniffing dogs.
This is similar to how single images of faces can be animated with GANs now.
You hit the nail on the head. Any system that hinders law enforcement’s assumptions and hunches will be touted as defective, anything that confirms those assumptions will be exalted.
This is why narcotics dogs are used. Studies show that dogs will signal upon trainer instruction, and that dogs aren’t accurate nor precise in practice. Such dogs exist to establish probable cause, not to determine whether or not a suspect possesses narcotics.
Several states have had success in identifying people with multiple drivers licenses using facial recognition systems on their photosets. Of course they are using DMV photos that are taken in a consistent way.
FBI or ICE could very well have a feed of identification photos from foreign countries that are close enough to get good success.
If the old photos are still on file, having multiple photos in the same style of a person a few years apart is actually probably quite a good training set.
On the other hand, governments with the power to mass surveil their citizens has proven to be a horrible idea.
Given the choice, I’ll take the lesser of two evils which is a company that is interested in knowing the things I buy and where I go, for the express purpose of selling better ads. But all the press goes to “let’s break up big tech”. I’d be much more interested in stopping the mass surveillance of citizens by an entity that has the power to kill, imprison, subjugate and arrest, rather than an entity that has the power to target me with ads.
In particular, they have to turn this data over to law enforcement by law.
Separately, law enforcement routinely breaks the law and steals the backend databases these companies gather.
The second is as a peer comment mentioned: information these companies collect is subject to end up in government hands anyhow. People are so quick to forget about the Snowden leaks of things such as PRISM: [Program participation entailed NSA access to] "extensive, in-depth surveillance on live communications and stored information." Examples included email, video and voice chat, videos, photos, voice-over-IP chats, and more. [1] Some of the companies confirmed to be working with PRISM are: Google, Apple, Facebook, and Microsoft.
I think things like this fall of people's radar because it doesn't get mentioned in the media. But it doesn't get mentioned in the media for the same reason it was never mentioned prior to Snowden either. These programs have not gone anywhere and, if anything, have grown only more expansive. The NSA didn't set up an exobyte scale data storage center [2] on a whim. Everything you submit to a major corporation in the US is something you should equate to submitting to the US government as well.
[1] - https://en.wikipedia.org/wiki/PRISM_(surveillance_program)
The choice is the crux of it. I'm not on Facebook, I choose not to have a "smartphone", I don't use free email or other web services that harvest my information and I avoid a variety of other activities that may offer every day conveniences at the sacrifice of my privacy to various corporations. I don't have this choice when it comes to the government.
We can debate exactly where the 4th amendment applies to prevent the government from hoovering up all our data so that they can track us all individually, in real time, and monitor all of our communications and interactions - but nobody can reasonably argue that the 4th amendment doesn't kick in at some point to prohibit this. Its always harder to claw back power the government has claimed than to stop them from claiming new ones. Its also the reason we need more transparency, so that we know exactly what information the government is harvesting and how closely we are being monitored and tracked.
I would be much more concerned if law enforcement were, for example, using these photos to profile drivers by race or using them in an otherwise illegal manner. That doesn’t seem to be the case here.
I think in the age of automated large scale image recognition we need to revise the concept of privacy. I don't think we should lose our privacy completely as soon as we leave our home. And with devices like the Echo we don't even have privacy at home. There need to be restrictions on mass processing of this kind of data. Otherwise we'll create a version of the world of "1984" where nobody can escape surveillance.
At least we can opt-out by not buying an Echo.
People can't choose to not have a face when they want to go outside.
Would you say that someone stalking you all day every day is alright if it's in public?
(Also, you seem to have missed that the police aren't using photos from just road cameras, but that's beside the argument really.)
Vehicles and roads don't enter the equation apart from people having gotten driver licenses. The licenses just make the largest database of US citizens' faces.
> [...] of dubious legality because it's a biometric search and should be done in a criminal case [...]
That is not the law. You have been misled, perhaps as the article's author intended. The author's feelings on whether such searches should be legal are not the same as democratically-enacted laws on the matter.
False. Until the observer effect occurs, I - essentially - do not exist but as a record in a database. Someone who knows me, my car, or has the ability to ascertain who I am from identifiers on the vehicle, is able to collapse that sense of privacy but, until that happens, I am just one of the many of the nameless mass. In that, I have privacy and you would be hard-pressed to prove otherwise.
>Tinted windows which obscure the driver’s face are, to the best of my knowledge, illegal in most of the U.S.
False. Tinted windows which obscure the driver's view is illegal. It's perfectly legal for you to have a high iridescent finish on the outside of your tint, which will obstruct the outside view of the driver (in sunlight).
>...or using them in an otherwise illegal manner.
When - in the history of ever - has law enforcement never abused the resources afforded to them? Your credulous, at best, belief in law enforcement's use of the system largely ignores the prevailing example given in the article - which was that they used the system on someone who was under "suspicious circumstance". The "suspicious circumstance" bar is so low that even a two-dimensional being couldn't limbo under it.
Some might argue that since the agreements were made that the three-lettered agencies would only use them for criminal investigations, that the example of the "suspicious circumstance" that was given just now is in fact illegal.
Law enforcement will always be a compromise. There will always be some who abuse it, and there will always be a need for it (if you want a reasonable level of civilization anyway). It's not binary. It's not "Give up all freedom and rights for infinite security" vs "Give up absolutely nothing for absolute freedom" with nothing in between. The line has to be drawn, but it is arbitrary, and a lot of people have different opinions on where it should be drawn (rightly so! there's not an obvious place to draw it).
We certainly have to be careful not to draw it too far on one side, but that doesn't mean it has to be completely on the other side.
Eg: I personally think it's too far to put surveillance cameras everywhere, but I'd be ok with cameras and facial recognitions on some major roads (they already have pictures associated with driver's licenses and various other documents for acceptable reasons).
Yeah, it might get abused sometimes, but everything can be. Everyone for themselves and hope for the best hasn't historically worked out so hot either. Law enforcement being completely neutered with zero tools and powers isn't very effective.
The problem with the compromise argument is that it only ever goes one way. We had certain law enforcement capabilities in 1965 and civilization didn't collapse. Why would we expect it to collapse if they had exactly the same capabilities today?
Whenever this argument is used, it's always to add new invasions. Databases that were never needed before, facial recognition that was never needed before. Why are they suddenly needed now, just because we can? Moving in only one direction over time isn't balance, it's marching toward a cliff. Meanwhile anything that does improve privacy, like encryption, is used as an excuse for new police powers as well.
It isn't necessary for law enforcement to catch everybody. And they wont anyway. Which is fine, because 99% of their purpose is deterring people from committing serious crimes, which they can do well enough without any fancy new technology.
You don't actually have to catch fugitives as long as being a fugitive ruins your life sufficiently that hardly anybody is willing to do it.
If police were limited to 1965 tech, you could see any police car coming miles away due to radar and radio detectors.
That's not true. Most states have a limit on reflectiveness. I checked Alabama and Wyoming (first and last alphabetically), and both specifically limit you to no more than 20% reflectivity by statute.
If drivers licensing agencies are allowing my records to be used against my interests (being returned as a potential match is certainly against my interest), it makes me consider the value of being a licensed driver.
I consider anything that encourages unlicensed driving to be a big problem -- unlicensed drivers are likely to be less well trained and less safe on the road, and are less likely to have evidence of financial responsibility (eg insurance) and may not be able to pay a judgement, and likely will drive unlicensed vehicles as well, so may be very difficult to track down in case of a hit and run.
Fundamentally --- drivers have consented to their photos being collected for a specific purpose, and using it for other purposes against the interest of the driver is a big problem.
I think that there were plenty of problems with the old approaches to policing, too. It was and remains the case that there's no need for detection if you can manage to pin the crime on a conveniently available candidate, whether or not that person did it; and pinning the crime on whoever you've got is greatly assisted by brute-force techniques that have nothing to do with technology.
That's what I understand... and that is Exactly why I posted my original comment.
This particular reading of the 2nd Amendment seems to have been frozen in time since the drafting of the Constitution.
Well God forbid it ever comes to that, but there's probably 60 Million people with guns here, and most own more than one... so basically a gun for every man, woman and child; its a formidable defense against tyranny. I mean it's not like we won in Viet Nam, to name one armed conflict that didn't go our way.
No it isn't? I mean, I know it's a very common conservative fantasy, but if the federal government comes after you, they are doing so with tanks and helicopters, and your puny guns won't change the outcome. Vietnam is not a good analogy. Your average American is dumb, fat and undisciplined. The Viet Cong were none of those things.
This is aside from the fact that an armed populace only encourages the government's law enforcement agencies to arm themselves more, and increases tensions in every encounter because the LEO has to assume the target is armed.
Because our tanks and helicopters have done so well at suppressing a bunch of afghani sheepherders, right? I don't think you're really qualified to evaluate the ability of the US's rural population to adapt to guerrilla warfare.
In all seriousness, it's easy to use superior weaponry to subjugate a populace... if one of the win conditions is killing everyone in the population to be subjugated. As soon as you have to care about things like "mass killings" or "civilian casualties" or "are you sure this isn't genocide" that gets a lot harder. And I'm relatively sure that the people who make up the US military would care, particularly when it's people they identify with that they're exterminating.
What is that supposed to do against the domestic population? Would they use their tanks and planes to blow up their own bridges and cities? Those are weapons used to claim territory from other governments. They're not very useful in a guerilla war where you don't even know who the enemy is.
> Vietnam is not a good analogy. Your average American is dumb, fat and undisciplined. The Viet Cong were none of those things.
Where does that leave the American government if that's the population they draw their soldiers and law enforcement from and they're each only trained for a couple of months? What do you do when half your new recruits sign up with the intent to turn against you as soon as you've trained them and provided them with weapons and equipment?
Fighting a civil war is hard. The US government is well aware of that, hence the bread and circuses.
> This is aside from the fact that an armed populace only encourages the government's law enforcement agencies to arm themselves more, and increases tensions in every encounter because the LEO has to assume the target is armed.
The solution to which is to expressly prohibit them from doing so. Random sheriffs have zero need for military tanks and riot gear and they shouldn't even be allowed to touch it. You don't need a SWAT team to serve a warrant on a check forger. You don't need a SWAT team at all, because anything it could legitimately be used for is the rightful role of the national guard.
Just because the domestic population is armed doesn't mean even 1% of them would shoot a police officer. And the people who would, tend to be the people who acquire firearms unlawfully regardless.
If the problem is that violent drug dealers have illegally-obtained guns, you can't solve that by taking legally-obtained guns away from peaceful hill billies and women who just want to feel safe walking home at night.
To fight an insurrection of its own populace? This is what is being described.
This has precedent. See Tianamen, Soviet Union sending the tanks to its satellite republics, the Turkey failed coup.
The only time the federal government can be stopped is to wait 4 years and electing a new president. Allowing individuals to own AR-15 has no bearing on 'freedom' from federal attacks.
Right, so the government has all the tanks and bombs they need to blow up their own bridges and structures.
Insurgencies don't have separate infrastructure. They use yours. Blowing it up hurts you more than it hurts them.
> This has precedent. See Tianamen, Soviet Union sending the tanks to its satellite republics, the Turkey failed coup.
Rolling in tanks is purely an intimidation tactic. They're designed to be hard targets that can destroy hard targets, but insurgents don't have hard targets. They use secrecy rather than fortification. To kill them with a tank you would have to know where they are, but if you knew where they were then you could go arrest them rather than doing anything whatsoever with a tank.
And even the intimidation value can be one hell of a footgun. Everybody has seen the photos from Tienanmen square of the man standing in the way of the tanks. That kind of iconic imagery is a massive fiasco for government and a major PR win for the opposition.
> The only time the federal government can be stopped is to wait 4 years and electing a new president.
It's not just about overthrowing the government. It's about making oppressive policies more difficult to implement.
It makes it more dangerous for corrupt police to sneak around like criminals without announcing themselves or outright lynch people in the streets, because it leaves people more ability to defend themselves against petty local tyranny as well.
Also notice that there is a large racial disparity in who gun control laws restrict from having firearms.
> Allowing individuals to own AR-15 has no bearing on 'freedom' from federal attacks.
In the case of an outright insurgency it's not about defense, it's about offense. It's about capturing soft targets for their resources. Twenty guys with rifles may not be able to take on an army base, but they can take on car dealerships and hardware stores and chemical plants and then drive off with their stuff.
And the obsession with the AR-15 is wholly misplaced. It's used by some bad people because it's used by a lot of people in general. It's popular. That doesn't make it different in any meaningful way from most other rifles, and the attempts to distinguish it immediately devolve into cosmetic differences like whether it has a pistol grip or certain types of mounts for ancillary equipment.
Isn't this coming back to your point? If you are fighting an insurrection, and start acquiring resources by force, you are losing the will of the people. At some point, your insurrection is just thuggery, returning to the point that guns only benifit criminals.
> It makes it more dangerous for corrupt police to sneak around like criminals without announcing themselves or outright lynch people in the streets, because it leaves people more ability to defend themselves against petty local tyranny as well.
History shows the police murdering people in the street because they happen to have guns. But since they fire dozens of shots against sleeping targets, having guns for defence never helps, and gets you branded as cop killer.
> And the obsession with the AR-15 is wholly misplaced. It's used by some bad people because it's used by a lot of people in general. It's popular.
Completely agree, it's just an example. A gun is a gun.
We've asked you so many times to stop posting this and other kinds of flamebait to HN, and you've so often ignored our requests to stop, that I momentarily banned your account when I saw this.
On looking at your recent comment history, though, I saw that you mostly haven't been doing this lately (that's good), so I unbanned you. Please don't do it again, though, because if this becomes a pattern again there won't be much slack left to cut.
The 2nd Amendment doesn't grant a right to bear arms - it prohibits the Government from restricting the right.
Is it not authoritarianism if it's Our Guys?
You can classify anything as hate speech. Propose a restrictive immigration policy? That's hate speech against immigrant populations, an attack on their economic opportunity. Propose a permissive immigration policy? Hate speech against the native population, an attack on their cultural survival.
Anything that affects anyone can be construed as being intolerant of anyone it negatively affects. Since all significant policies affect people, anything your opponents say can be classified as hate speech by an authoritarian who doesn't like them.
Isn't this the argument that the concentration camps being run to detain refugees and separate their families aren't authoritarian because the government and their media allies say so?
The problem with the 2nd amendment approach is that a lot of the militia types are quite in favour of authoritarianism so long as it hits people they don't like.
Only if you're the government and their media allies.
Anything that puts people in cages is authoritarian. You can argue that sometimes it's worth the cost, but then you're arguing that sometimes authoritarianism is worth the cost. And generally speaking it isn't and we should have a high bar for all imprisonment.
(Though the immigration case is somewhat confounded if being in the cage is only an optional alternative to being immediately delivered out of the country and released there pending an immigration status decision.)
> The problem with the 2nd amendment approach is that a lot of the militia types are quite in favour of authoritarianism so long as it hits people they don't like.
So what's stopping the people who don't like that form of authoritarianism from taking advantage of the 2nd amendment themselves?
It's only six dead children so far. Barely moves the outrage needle. https://www.theguardian.com/us-news/2019/may/23/migrant-chil...
https://www.aclu.org/issues/privacy-technology/national-id/r...
Some states listened and intentionally dragged their feet. Mine is one - we needed an extension in order to fly -- but we're rolling it out here now too.
Funny you should mention that in your call for caution against black pilling. When SS was proposed there was massive popular outcry against serializing citizens with a federally assigned number. The USG assured everybody that SSNs would only ever be used for the distribution of benefits...
Everyone is still being spied upon, only now people seem to have a sense of inevitability about it.
The German government became very interested in open source firmware, I've heard talk of RFPs requiring it from vendors - which compels vendors to put pressure on manufacturers, leading to fewer binary blobs.
Google became very interested in the power processor. Years ago their decommissioning process involved a hole being put through the Intel processor. For a long time I just assumed that there was a deal in place with Intel requiring a certified field destroy, but I'm now thinking that it was their distrust of Intel's backdooring. That may sound extreme, but nobody can say with certainty that whatever bs is occurring in ring -6 isn't caching private keys. Intel is now trying very hard to fool people into believing that they're going to be open sourcing their firmware - this will not happen.
As far as the social consequences, that is hard to say - I'm less plugged into popular culture than most. I'd be surprised if it didn't influence people's views of the government though, especially if they had voted for Hope and Change™. Greenwald timed and ordered the releases perfectly so that the government would predictably lie, only to be proven a liar the following news cycle. This pattern repeated a few times before they got the hint and STFU. And then there was the director of the NSA showing up at Defcon in a black t-shirt and jeans...
But besides arguing there are things you can do. As coders and techies we can contribute code. We can encourage open source and make our own code open source. We can help others find secure forms of communication. We can also do what many others in the public can do: protest, write our congressmen, and raise hell. How much and how little of this you want to do or have time for is okay. But I wouldn't say that just because you haven't convinced others means you shouldn't try. It just means you should change strategies. You after all, no matter how smart or right you are, are not perfect and neither is the person you're trying to convince.
« Rep. Jim Jordan (Ohio), the House Oversight Committee’s ranking Republican, seemed particularly incensed during a hearing into the technology last month at the use of driver’s license photos in federal facial-recognition searches without the approval of state legislators or individual license holders. »