Schneier on Security
A blog covering security and security technology.
« Renew Your Passport Now! |
| More on the HP Board Spying Scandal »
September 18, 2006
Does this EyeCheck device sound like anything other than snake oil:
The device looks like binoculars, and in seconds it scans an individuals pupils to detect a problem.
"They'll be able to tell if they're on drugs, and what kind, whether marijuana, cocaine, or alcohol. Or even in the case of a tractor trailer driver, is he too tired to drive his rig?" said Ohio County Sheriff Tom Burgoyne.
The device can also detect abnormalities from chemical and biological effects, as well as natural disasters.
Here's the company. The device is called a pupillometer, and "uses patented technologies to deliver reliable pupil measurements in less than five minutes for the detection of drugs and fatigue." And despite what the article implied, the device doesn't do this at a distance.
I'm not impressed with the research, but this is not my area of expertise. Anyone?
Posted on September 18, 2006 at 1:39 PM
• 50 Comments
To receive these entries once a month by e-mail, sign up for the Crypto-Gram Newsletter.
Cops use peripheral light sources to monitor pupillary tracking as part of a simple first-cut drug test. Maybe this is something like that.
> in seconds it scans ... pupils to detect
> a problem.
Wow. I'll have to use that on my students.
Wow! This thing can look in my eyes and tell if I've been in a hurricane, earthquake, or flood?
Where do I get one?! :P
I especially love the "and what kind" bit. Okay, I buy that drugs will interfere with pupil dilation, but I refuse to believe that each drug has a specific enough reaction in every person to allow for identification.
>The device looks like binoculars, and in seconds it scans an individuals pupils to detect a problem.
> uses patented technologies to deliver reliable pupil measurements in less than five minutes
...so which is it? Granted few seconds is technically less than 5 minutes. I guess it might time the response rate of pupils and come up with something, not a doctor so I have no idea what I'm talking about...
Since "eyes are the window to the soul", will it also detect heartless bastards?
There is a whole history of quack medicine that claims that diseases etc can be plainly identified by looking at eyes (a very few actually can be).
This looks like more of the same.
Wasn't there some French "crime expert" back in the early 1800's who used measurements of the eyes, skull, fingers, etc., to determine if someone had crimminal tendencies?
@Realist: You're probably thinking of Franz Joseph Gall and phrenology?
Shura, I believe Realist is referring to iridology.
@Realist: You're thinking of phreneology (sp?) the study of bumps in people's skull to determine if they are criminals or not.
Oh, my bad, that was jayh that was talking about iridology.
IIRC, changes in pupil diameter are also indicative of physical attraction (that is, if we're looking at someone we find interesting or attractive, there's a slight increase in our pupil size), as well as recognition (is this your card?).
Disparate pupillary size can be caused by head trauma. And even with the counteragent, someone who has gone to the optometrist's may have enlarged pupils for several hours afterward.
Sounds like snake oil to me, or at least a gizmo that functions only within a narrow range.
And any decent criminal lawyer will consider calling the system into question - if it mistakenly identifies someone as intoxicated, then it can't possibly have correctly assessed the defendant.
And in the hands of Harrison Ford it can suss out replicants.
curses, k. I wanted to make a replicant joke.
(that device was called the Voigt Kampff Test - I wonder if we can bust their patent by demonstrating Bladerunner as prior art?)
Yes, this can work, sort of, but only with cooperation from the defendant. As a Vision Scientist, I would expect it to have a range of uses, but I am very doubtful that a Police Officer would have the skills to operate it, even if it's largely automatic, and I am sceptical about some of the claims, particularly the one about detecting tiredness! However, it will doubtless detect and categorise some drug effects, but such evidence would never stand up in court, not on it’s own, and there are many medical conditions that would confuse it. What happens if a defendant has a condition like anisocoria (difference in pupil size – many different natural causes) and they get a ‘fail’ light? A night in the cells perhaps? That would be indefensible, and grounds for legal action. I think that it will help spread eye infections too, unless wiped down with disinfectant each time.
The pupils of one's eye act like the apeture of a camera: they control depth of field (primarily) while having the secondary effect of altering the brightness of the light hitting the retina. Pupils contract when the eyes are exposed to bright light (relative to whatever the previous condition was) only as a stop-gap while the eyes re-configure themselves for more available light. This reflex also happens to prevent the retina from being burned from over exposure. It normally happens in both eyes simultaneously, so long as the efferent fibers from the optical motor-related nerves are intact--even if one eye is blind. More than a few drugs, sources of trauma, and illnesses can upset the function of this reflex.
If you want information about one's state of conciousness you are going to have to rely on more than just the state of one's pupils--everything else should go to hell first, if that is what you are worried about.
On top of all of this, people's eyes should normally dart about in a damn-near random seeming fashion even during normal conversation. It is only if the eyes of the person you are talking with _really_aren't_moving_ that you should perhaps even notice something out of the ordinary.
So, does this thing work as advertised? NO. Can it tell what somebody is on? NO. (It often takes a blood test to figure that out conclusively.) Is it a waste of time and money (perhaps so that somebody can feed their eyelash fetish)? YES.
Oh, and while we are at it, imagine the trouble that something like this will cause for people with one glass eye.
(The great things that can be learned from some introductory neurology and a first responder course.....)
I imagine that it is a real trip to work in the QA department that tested this device. Really.
Within five minutes? Isn't that enough time for the cop to observe behavior and suspect a problem?
I was thinking along similar lines when the title came over RSS.
Actually, pupil response to gross stimuli (like a bright light) and whether they respond symmetrically is one of the evaluations the EMTs make should the "victim" be unconscious, and is, in fact, one of the ways to check for brain damage or a stroke.
I suspect, assuming a "good" baseline (I can imagine being a worker on the project, taking puffs of pot to calibrate the technique, then using other illicit drugs as part of the testing/calibration regime, might be enjoyable for some) that _something_ useful FOR THAT SUBJECT might actually be discovered. This assumes the instrument is not as interested in the actual diameter of the two pupils but how well they respond to suddenly changing lighting levels, of course.
The funny thing is that I'm not sure such a test could _really_ differentiate between just plain old fatique and the use of legal (or illegal) drugs. Note that pupil dilation _is_ a side defect of various medications.
The hell of it is that one set of baselines won't extend well to the population at large. I've seen it said that humanity is a study of exceptions, rather than rules.
I can understand the desire for a relatively "non-intrusive" way to test for illicit drugs in someone's system... or as a lie detection, by asking leading questions while the subject is looking into the device, but, hey, it's not much different from a breathalyzer.
Wanna bet this will be useless once it's in the field and has an awful "false positive" rate?
> Wasn't there some French "crime expert"
> back in the early 1800's who used
> measurements of the eyes, skull, fingers,
> etc., to determine if someone had crimminal
You're probably thinking of the Italian criminologist, Cesare Lombroso.
There was also a French biometrics researcher, Alphonse Bertillon, but his research was aimed at identification (before fingerprinting), not identifying criminal tendencies.
Here's another story, with a rather different claim:
"The machine does not show whether a person is using drugs, only whether their nervous system is impaired from something such as fatigue, an illegal drug or an over-the-counter medication. If the EyeCheck shows that someone is impaired, a urine test then is administered to determine what caused the impairment."
If memory serves, standard police procedures include some equally as quick and far less expensive ways to determine that someone's nervous system *might* be impaired.
I hope that they never use this on me. My eyes dialate independantly. (Always have. I have to warn EMTs about it or else they tend to think I have a concussion.)
My bogocity meter is pegging on this one. Next thing they will bring in dowsing to find the driver's stash.
Another pseudoscience needs to be applied here.
Phrenology is the science of determining a person's personality by examining the bumps on their head.
Phrenotherapy is changing a person's personality by changing the bumps on their head.
Don't do that! *whack* Don't do that! *whack*
Thanks, Ken. Those were the people I was thinking of... I remember reading about this way, way, way back in my teens... I'd forgoten that it was an Italian who "startedd" the notion, but recalled reading that a French police cheif had tried to apply the theory in a particular murder case to prove the suspect was pre-disposed to being a murderer...
Several times a year, a chronic injury flares up and my pupils look like I could be on a combination of hash, and even more dangerous stuff like meth, crack, tobacco and alcohol.
If this kind of thing came to be used, I'd be under house arrest for a week at a time, 1-3 times a year.
"Is this one of those tests?" -- "Rachael", Bladerunner
"Capillary dilation .. the so-called blush response? Fluctuation of the pupil? Involuntary dilation of the iris?" -- "Dr Tyrell", Bladrunner ("I want to see a negative, before I provide you with a positive.")
The technology sounds plausible, pupils dilate and contract at different rates depending on drugs, and have different ambient dilation.
Probe with a range of light intensities and measure reaction time and amount. Simple enough.
And, yes, different drugs have different types of effect, LSD makes your pupils dilate way wider than normal...
Brain state affects also the amount of jitter or what-have-you in the direction of the person's gaze, which is easy to track.
It might also look at "capillary dilation".
Use ir to sense the pupil direction and dilation, and visible light to stimulate responses.
The detector might not notice small amounts of drug, perhaps only amounts that are relevant, but that would normally be considered a plus. It might also notice non-drug-related brain mishaps, but that might also be considered a plus. Depends if you are just a cop looking for a bust, or someone who actually cares about what is going on.
Even cops would probably find it useful to have something like this to find more busts, and yes, I think this could be done at a distance, if the environmental lighting is taked into account, and better if it is optimised.
Technology to track people's gaze using ir, for use in directing a cursor has existed since the late 80s, here is an example (not sure if this uses ir or something else)
Think this is new, I don't. And I guess flash 'special' cops (intelligence agents, feds, contractors, specialists) have had this for ages.
No, it would *not* stand up in court. It would approach accuracy only circumstances where a blood sample is available and more reliable, so no court should accept this as anything but an indication for discussion of what might be going on in a particular place-and-time.
@Alan "My eyes dialate independantly."
Afaik this is normal .. I guess your pupils dilate to different extents, or at different rates, to each other. I'm not a doctor, but I vaguelly recall having this test explained once when I was concussed ;)
Cops approaching a violent crime, btw, work pretty much on a 'default deny' policy, judging the client's potential behavioural reactions "conservatively", which is only sensible.
They then try to make lists of known exceptions, and share that information around, because after all, even in the worst legal systems, their job is to care for the innocent.
They also try to get as many indicators as possible, and this is liekly to become more and more automated, as we've seen. Various types of automated screening aggregated to give a profile described in terms of a small number of words and numbers -- easy to understand in heat of moment.
So those cases such as people with a natural asymmetry in their dilation, the cops are going to be aware that the device is not 100%, and they will use other factors as well. But the job is inherently one that encourages excessive reactions and prejudice, for the reasons described above.
I don't see that this invention is bad if used by police defensively, but I can see how it is bad when many things like this become prevalent throughout society, and especially where the laws and policiing practices are not altered to reflect the greater ease at detecting a range of what were previously universally classed as "infringements".
After having worked on a volunteer fire dept for a couple years, you start to be able to very quickly sort people into the correct bins:
- altered via drugs
- altered via concussion
- altered via stroke
But it takes compiling a bunch of individual parts together. Motor control vs. general alertness vs. eye reaction vs. speech, etc.
Eyes are only a part, although they're a pretty good indicator of *something* being wrong. The first thing a medic/emt does on scene with a patient (who isn't obviously a trauma/medical case of a more serious sort), is start asking questions to see how alert and oriented the person is (name, age, location, what happened), then a pen-light to the eyes and you've got a very good summary of what's going on.
This is the kind of processing that Malcolm Gladwell talks about in his book "Blink". And a good cop/emt/ff learns to identify the overall signature, composed of many individual details, and you can pretty quickly sort people into what kind of a problem they are having.
And does it work with someone that's only minorly impaired? yeah, it does. I actually find that major impairments are more puzzling. Are they just *that* plowed, or do they have a concussion as well? (drunk and fallen and not very responsive).
I honestly don't think a machine would be abe to do this, not unless it was monitoring people for a very large number of factors, and looking for the subtle hints that add up the real issue.
So ... if it *could* tell whether I had been taking marijuana, cocaine, or alcohol, could it tell if I had been taking combinations of these substances ?
As an optometry student near Portland, OR I got to observe Portland PD Drug Recognition Expert training during the Rose Festival. (They pull the druggies off the street and ask them if they would rather participate as a subject in DRE training or go to jail.)
The officers measure pupil size in low light and direct light conditions. But these tests also have to be performed in conjunction with other tests like horizontal gaze nystagmus detection, counting to 30 seconds with their eyes closed, ability to follow commands, ability to walk a straight line, and more.
By performing all of these tests (not just pupil size testing) they can detect the exact drug class the perp is using (cannabis, amphetamine, barbiturate, heroin, cocaine, etc).
EyeCheck supports their claim that it detects fatigue from a study they cite about rate of pupil constriction and rebound (reflex amplitude). Observation of rate of constriction would be variable from officer to officer and could only be described as brisk or sluggish, so computerized, objective, accurate measurements would be nice for pupil constriction velocity and reflex amplitude.
So the fancy machine apparently does help with fatigue detection. It also gives a nice report on pupil size measurements, so you couldn’t accuse the officer of just making stuff up. If a LEO had to work with a judge who was highly critical of how reliable the officer’s data is, then this machine could helpful. However, if the law enforcement agency has never had a problem with a judge, then I would find it hard to believe they would pay extra money for a fancy "pupillometer" when a simple penlight and pupil size card does the same thing, except for “fatigue detection.��?
Samh> So ... if it *could* tell whether I had been taking marijuana, cocaine, or alcohol, could it tell if I had been taking combinations of these substances ?
That question suggests a fairly straightforward evasion strategy. :^)
Perhaps it's not looking at dilation, but anylizing the blood which is very close to the surface in the eyball. Similar to the plug/clamp that is placed on someones thumb in hospital to get blood cell counts.
Sounds like this would be of limited usefulness, even if it were remotely reliable.
As I understand, pupils change depending on how much you personally like what you're looking at. They also change depending on how happy you are and what you are thinking. You could dialate your pupils just by thinking of showering with your favorite movie star. Too many subjective variables here.
It sounds like this is an attempt to test a deviation from a baseline and that they know what the baseline is. I had an eye injury and no longer see from one eye - my left pupil does not dilate much at all, even in the brightest light. My doctor told me that I should mention this to anyone treating me if I'm ever in an accident so they don't conclude that I've had a brain injury (or am drunk, stoned, tired, etc.). I'd like to know that a technology like this is able to deal with exceptions like these. One size rarely fits all.
"If memory serves, standard police procedures include some equally as quick and far less expensive ways to determine that someone's nervous system *might* be impaired."
I suspect one reason there's a push to come up with machines to do this is to enforce some "objective" means in order to allow *any* yahoo to be a member of the security forces. If you equip your thugs/minions with a machine to do the real thinking, you can get some *real* sheeple out there to wear the jack-boots for you and not depend upon individual intelligence or initiative... because, if you *do* have to depend upon intelligent individuals, you may have to contend with free-thinkers who won't follow orders blindly.
Wow. Maybe I should write something like this into one of my stories?
As I am not aware of any medical uses for this type of technology, I am therefor inclined to think that it is not very reliable (police Dept's are not awash with money when compared to privately funded hospitals).
There may be something in the device, but if you read the only online "medical research" paper quoted (Forensics Drug Study Author: Jack Richman Affiliation: The New England College of Optometry 2002). They compared urine screaning results with the responses of the pupil in 146 test subjects of which only 126 where usable (16% of the population rulled out to start with).
You read the following,
"Pupillometer and urine screen for a variety of drugs. Seventeen percent were considered false positives on the Eye Check™ and three percent were considered false negative.
Based on this,
The Eye Check™ appears to differentiate subjects who were potentially drug impaired from normal in the majority of cases. The sensitivity was 86.2% and the specificity was 78.85%."
Which is not that good on it's own, but drops down a lot when you compare the results with the original population size (ie it is right maybe 2 times in 3). The paper goes on to confirm this by saying,
"Using the pupillometer solely, the Probation Department could have saved more than half of the drug test expenditure."
Which basicaly means that the device is not very reliable and is an "indicator" not a "test" (otherwise the saving would have been up in the 90% range).
It's kind of like sniffing somebodies breath for alchol, it's a reasonable indicator but you are not going to make a real bet on it.
One thing not mentioned is what else the device could with a little modification do (afterall the optics / computer etc have been designed).
For instance anybody who has been to Hospital in recent times will almost certainly have had their oxygen saturation (SpO2) and temprature measured by the use of infra-red electronics (redlight clip on your index finger for Sp02 and opticaly down your ear for temprature). Both these tests are very fast and virtually real time which is very usefull for ER doctors.
The human eye does not repond well to majority of light wavelengths that the lens is transparent to, and in a dark environment the iris will open wide over a period of time (upto 20mins).
Due to the nature of the retina which amongst other things makes it possible to shine a beam of light in and measure the absorption of the blood (think red-eye in photographs). I think it would be possible to do an absorbtion spectrum of a persons blood through the eye.
What I don't known (not having seen/read any papers on it) is if it is going to be sufficiently wide to be usefull in all conditions (think cataracts and other lens / retina defects).
Also do various drugs have sufficiently recognisable absorbtion spectrums when disolved in the blood stream, again I do not know, but being organic chemicals there is a reasonably good chance they do.
Expect this technology to be re-branded and re-deployed in the near future for use at airport checkins.
I'm sure homeland security is a much more profitable line these days.
This thing does a Horizontal Gaze Nystagmus test. I was trained in it when I was a probation officer. The eyes have various responses based on what kind of drugs you use. Pupil dilation is one. The pupil of a person on a stimulant will contract when light is shone in it but will slowly begin to pulse and gradually dilate again even if the light remains. Stimulants also cause the eyeball to bounce when the gaze is focused to one horizontal extreme or the other. The eye can't hold the position and the pupil bounces back and forth like a pinball. Even mild stimulants like caffeine can cause this effect. Depressants cause other effects such as sluggish contraction and dialation.
I thought it was all BS until I took the training and started using what I learned on my probationers. It really does work. In fact, there's a scene in the film Scary Movie where some dark haired girl is sitting on a bed talking to a guy. Watch her eyes during that scene and you can see her pupils bouncing back and forth. She was probably high on coke or meth. I pointed it out to my wife and it freaked her out. The person can't feel their eyeball bouncing around but you can plainly see it happening.
This is one creepy device. There seem to be no adjustments for hand tremor (either for the tester or testee) nor is there any indication that the device is protected from signals (EMI, sound, etc.) that could conceivably add noise to the (I assume light/motion sensor, not electronic) signals being input.
Furthermore, there are no published tests that distinguish between, say, an eyelash in my eye and an hours-old hit of crank.
Finally, John Daugman, in his extensive reviews of iris recognition concepts, makes a strong case for establishing an ideal crossover rate between the sensitivity of the measure and pass/fail decisions. Legally, false positives and false negatives (Type I and Type II errors, for you statistical folks) could lead to a whole lot of unfounded arrests and lawsuits against both the manufacturer and the state of West Virginia.
Shall we return, Bruce, to your oft-stated advice to those who would seek to protect us: nothing beats a trained human observer when seeking to predict/understand human behavior.
Presumably the device would only be used to screen for people who need more formal testing. You could level a lot of the same criticisms at the little bend-backward-touch-your-nose and other motor coordination tests, but they are just for screening.
@David Langford wrote "If a LEO had to work with a judge who was highly critical of how reliable the officer’s data is, then this machine could helpful."
I'm assuming this is to be used in the field in order to make arrests. Once they get an arrestee to the station they can take a blood test, which *should* be standard practice.
Out of interest, not so long ago police in the UK were testing a device that could potentially analyse whether or not someone had been taking illicit sustances by wiping their forehead with the device.
Note also that breathalyzers have failure modes as well. They can, for example, read high if you've used an alcohol-based mouthwash shortly before being tested.
I've never been in a DUI situation, but is the case that one is arrested *before* taking a blood test, based on breathalyzer or motor-coordination tests? Or are you simply asked if you will take a blood test and arrested only if it shows an illegal blood alcohol level?
just a variation on the old nystagmus test.
cop shines a light in your eyes, tells you to follow his finger. if you squint, he can't see your eyeballs.
another inroad into monitoring your body, which you can train yourself to defeat. 20 years from now, they'll be doing penile plethysmographs at traffic stops.
My feeling is that this is going to be used in the field to force a urine or blood test. I find myself concerned that this could be used more and more randomly. This is of personal concern to me (and will be to others like me) because I - completely legally - use both prescription amphetamines and narcotics for medical conditions. Obviously, I arrange not to drive if I must take a narcotic, but I don't know what small traces a blood test can pick up. I probably need to talk to the prescribing doctors about this issue to keep myself safe.
The main medical reason for measuring pupil size and reaction is to assess potential internal brain trauma. Pupil constriction is controlled by a nerve which comes from deep within the brain. Swelling of the brain can compress this nerve and impair constriction - hence if someones pupils are "fixed and dilated" they are in trouble. Accurately monitoring a patient's pupil size and reaction rate over time can provide a good indicator of brain condition.
As for drug impairment testing - yes, barbituates can impair constriction but, without a series of baseline measurements under normal conditions (as mentioned elsewhere) a single spot reading is meaningless.
We use a Pupilometer in my lab. It takes a few seconds to calculate the diameter of the pupil, but nothing more than that. It doesn't look particularly like binoculars.
The guy at the top who said he'll use it on his students, is a scumbag.
Does the size of the person change the amount of effect the drug has?
Schneier.com is a personal website. Opinions expressed are not necessarily those of Co3 Systems, Inc.