DIY Iris Scanning?

gadzook33 asks: "There have been rumors floating around about DIY iris scanning, using digital cameras for biometric security. Iris scanning presents a fantastic alternative to password-based authentication but hasn't really come to our desktops yet. I've looked around but can't find any concrete material on the subject. Is anyone doing this? Are there any efforts to develop open software for this sort of thing? Are patents holding things up? Given that passwords are an almost defunct technique for protecting data in certain situations, it would be nice to have an alternative."

Comments

eye scanner

[indy_Muad'Dib]

it seems to be quite possible with a very high resolution camera, something +4mpixels

Why would I want to...

[YowzaTheYuzzum]

... give anyone an incentive to gouge out my eyeballs?

Better than retina scanning

[jamesh]

which is where you use a laser to illuminate the back of the eye, and a camera to take a picture of the illuminated retina and then use some sophisticated pattern matching to recognize the unique pattern of scars left by previous scans.

Re:Better than retina scanning

[fyonn]

Retina scans are not likely to injure you, but are considered less acceptable than iris scanning as it gives away too much information. Yes, it can uniquely identify you, but it can also divulge various health issues and show if you're pregnant. This is usually information that employees prefer to reveal on their own, than have the door security guard congratulate or commiserate them about information that they haven't told their partner about yet, or might not even know them selves.

dave

Re:Better than retina scanning

[fyonn]

and of course, it could also be used to affect your future career or health insurance as it reveals to your company things such as high blood pressure, diabetes, drugs use and leukemia. hell, even aids and syphilis.

things most people would rather keep to themselves.

dave

Re:Don't Look!

[rts008]

RTFM, n00b!
Oh, and carry a few spare eyeballs with you- you know...backups!

really?

[Unknown_monkey]

Out of all the things to DIY, what would drive you towards a DIY project involving possibly lasers or bright LEDs and your eyes? Some things you shouldn't go bargain on, like never buy the cheap toiler paper. For both my eyes and my brown eye, I think it's worth spending the cash for premium.

Re:really?

[jamesh]

This iris is the front part of the eye (See here). No need for any special sort of illumination above a light bulb. The Iris Recognition article on wikipedia is also somewhat informative, it even mentions the problem of cosmetic contact lenses.

Not an alternative...

[Zadaz]

Iris scanning presents a fantastic alternative to password-based authentication...

This is an all too common mistake about biometrics. Security should never rely solely on biometric identification. Unlike a password or a physical key, your biometric information can't be changed. Which is its strength, right? No one can change their fingerprint to match yours!

However, any system can be spoofed or cracked. And if someone figures out how to feed information into a scanner that looks (to it) exactly like my iris, well then I'm fucked. That person is me anywhere they do an iris scan.

It would be like someone stealing your passwords and you not being able to change them.

Useful? Yes. But as an additional level of security, not an alternative.

Re:Not an alternative...

[ryanhos]

There is merit in your argument. The basic idea of using biometics as a an additional level of security is unimpeachable. However, you miss two key issues in play here.

1.) A key and irreplaceable component of any authentication instrument is a revocation feature. You state that biometric passwords are not changeable. Biometrics are just as revokable as passwords. In both cases, the user must recognize that the instrument has been compromised and tell his keymaster. In the password case, you simply change the password. In the biometric case, you must remove that instrument from the authentication whitelist. I think what you're trying to get at here is that biometrics, once revoked, cannot be resued. However, your statement: "That person is me anywhere they do an iris scan." is false because the biometric instrument can be revoked. That person is you everywhere they use a password until that is revoked also....

2.) Biometrics are growing up. Soon the days of simple image processing will be gone. Additional checks like measuring pupil reaction time, eye movement characteristics, blink characteristics, eye pressure, peripheral vision, visual acuity, etc will be implemented. (Press the button when the line-drawing is in focus (using a new image and different focus path each time.); Press the button when the spot comes into your peripheral vision (same randomness, etc.)) Varying the algorithm by which the image processing software recognizes the eye will also harden iris scanning. All of this works to make biometrics an increasingly attractive ADDITIONAL layer of security. I've always been told that the best security check involves two things: something you have, and something you know. Biometrics can work quite well for the "something you have." part.

Re:Not an alternative...

[Vo0k]

"That person is me anywhere they do an iris scan." is true. Except they get "Sorry, you don't have a clearance" as a reply, just like you do. If a building security is based on iris scan, sure they won't be able to enter after your iris pattern is revoked, but so won't you. Meaning no entry to the building, sorry sir, you must look for a job elsewhere, at least till we update our security system.

As for 2), the basic feature of biometrics is that it's simple. You touch a surface or look into a lens, and that's all, no typing passwords, no entering codes or searching your wallet for magnetic card. Take it away and you take away half of the charm of biometrics. You only leave the scare "they will knock you out and take your eye out in a dark backstreet to break in" plus vague and unreliable info about high security, which is neither verifiable nor unhackable and definitely doesn't appeal to management.

It's a bumpy road ahead of biometrics.

Re:Not an alternative...

[Zadaz]

1) Sure, my biometric permissions are revocable, but not re-issuable. At least no security outfit in their right mind would reinstate your biometric print once it had been broken.

If simple biometrics become prevalent, then someone stealing my iris print (for example) would pretty much end my life. I wouldn't be able to have a bank account or any other kind of security. Either my accounts would be wide open to whoever had a copy, or no bank would issue an account to a security risk.

At least until I could grow a new eye. It's identity theft on a very personal level.

2) Sure, they're getting more advanced. They could hardly be more primitive. However there are two problems with making them more sophisticated:
a) You can't make security so sophisticated it can't be broken. (duh.)
b) The more complex a system is the more likely it is to fail. I'm not an expert in the field, but many of the things you propose would ilkley prevent me from accessing my account if I was ill or under the effect of any number of legal drugs. Which is of course unacceptable.

A system that sophisticated will cost a ton of money. Compare that to to the cost of a card reader and 12 button keypad found on most ATMs. The amount of ATM fraud based on stealing user ID's at the terminal is much smaller than cost of installing and maintaining biometric devices and will be for the foreseeable future.

Better Yet, Go Multi-Tier

[DumbSwede]

I personally would like to see multi-tier biometric authentication built into the OS. Log on with a password and a finger scan; any File I/O challenge with voice recognition; visit a secure site, submit to iris scan. Mix it up, occasionally challenge with authentication questions when actions seem either dangerous (downloading executables) or deviate from usual usage patterns. How aggressive to be in challenging for authentication and what types should be settable by the user. This kind of thing might be very useful in keeping your teenage kids from downloading Kazza like malware on your family computer, not just keeping your computer secure from crooks and spys.

Re:Better Yet, Go Multi-Tier

[corychristison]

Or better yet, you could have a biometric finger-scanning keyboard.

Each key will be a finger-print scanner. If you type home-row then match the keystrokes to fingers that are supposed to be used. When you press down on a key, it will take a photo/whatever of your finger and analyze it to a small database at the system[/kernel?] level to see if they match up.

If the finger print doesn't match, keypress is ignored.

You would need some incredibly quick fingerscanners for each key, and it would be VERY pricey. It would also force you to use home-row, too! :-)

Possibly incorportate this into the mouse, as well? ... somehow

Yeah... I need sleep. But it's a thought. Would be VERY hard to crack, esspecially at the kernel level. You wouldn't be able to just swap out keyboards.

Aaargh! Wrong laser!

[myowntrueself]

I thought I told you to label those buttons, Emory?

Re:useless

[Invidious]

No, they can't. In order to get a good image of someone's iris, particularly without being noticed, you'd need a long (at least 400mm) macro (true macro, 1:1 reproduction) lens with very, very little sperhical abberation or chromatic abberation, and very sharp to boot. And, likely, a buttload of light shining into the person's eye. Since you can't get a lens like that, and it'd be monstrously huge anyway, it's not a problem.

Static Iris Scanning is useless, not dynamic

[iendedi]

While it is true that one could hold up a photo of your iris to a camera and spoof a static iris scanner, doing the same to a dynamic scanner is not practical.

What is a dynamic iris scanner? One that looks not only for the unique patterns of the eye, but also simultaneously measures retinal response to stimuli such as dimming and brightening of the display. This is much more difficult to spoof (you would essentially need to build a model of the target's eye that could respond to external stimuli and then hold that up to the scanner).

Combined with facial recognition, dynamic iris scanning is very potent. First it recognizes your face and then your eye and then the retinal response with stimuli that is timed to be somewhat random. Just don't try to log on after a night of pubbing.

Standard warning:

[Strolls]

Do not look into laser with remaining eye.

Biometrics is very hard to get right.

[gweihir]

Face and Iris recognition have been fooled with printed pictures. Fingerprint sensors with $5 fakes. The list goes on. There is really not a lot of defenses available against this.

And you cannot change your face or iris, like you can change a password....

Little red lights

[G1975a]

Aren't the little red lights on the bottom of my mouse iris scanners? That's what they told me at work.

How I would do it...

[cr0sh]

I would start out with a cheap USB web camera. First, I would hack it in some manner to allow it to macro-focus. I would go down to goodwill or a pawn shop and pick up the cheapest, most busted VHS video camera being sold. From this I should be able to get much of the optical components and the needed eyecup.

I would attempt to obtain a fake eyeball of some sort. While it wouldn't work perfectly, it would give me some sort of method by which to focus with. Mounted with some tape to the eyecup, and then positioned in front of the webcam, I would be able to determine the focus fairly quickly.

I would then set up some kind of "ring illumination", wherein I would create a "ring" of LEDs - red/green/blue/IR - through which the webcam would peer. Focussing again might have to be adjusted. This ring would be set up in such a fashion so that I could trigger which set(s) of LED's would be active at once - likely via USB control, too.

Once I had that set up, and focussing correct, I would then work on the software. For this DIY project, I would simply set things up to take multiple image captures of my own eye, process the images through some filters to reduce the information to just my iris (cue on the white of eyeball, and black of the pupil), then (in some manner), use these images to create an "eigeniris" image, some kind of "average" of all the images I took (over several days or months, in different levels/conditions, so as to have the best average available). Then, the software could take an image, compare it to the "eigeniris", determine if it falls within range, and use that to trigger or deny access (to whatever).

That would be the route I would take if I was doing this. Overall, the hardware portion seems the simplest to implement - the software is where you will bog down. Just like any other pattern recognition project, I would imagine...