Slashdot Mirror
3D Holograms Detect Fake Signatures
Roland Piquepaille writes "Several sources reported last week that a new technique that produces 3D holograms of handwriting could be used to detect fake signatures on checks, credit card receipts or other important handwritten documents. Here are pointers to Nature, Scientific American or BBC News Online. Instead of using 2D techniques to look at the sequence of pen strokes in a signature, this new method is based on 3D micro-profilometry which permits to translate the writing into an image showing dips and furrows of the sample so that anomalies can be detected. If you plan to imitate your spouse's signature, beware! Forensics have a new and very efficient tool. As an example, for the use of ballpoint pens on normal paper, the success rate was 100%. You'll find more details, references and pictures in this overview."
Is my own writing that similar? What happens if I let my guard down and something slightly different, will I be arrested for fraud by forging my own name?
When I was at IBM yorktown Heights, the guy in the next lab over built a pen that had piezo acclerometers and pressure trnasducers built in. You got the time and pressure curves of the 2-d signature as it was signed. IBM never marketed it.
Some drink at the fountain of knowledge. Others just gargle.
Abstract
For legal purposes there is a requirement for the validation of signatures and handwritten documents. A helpful method in this respect is the so-called superposed strokes analysis, based on the observation of some characteristics in the writing, such as some letters and their dynamics.
This paper introduces a promising new technique for superposed strokes analysis based on conoscopic holography. Through a non-contact 3D measure a 3D profile is created of the superposed strokes that allows the writing dynamics to be determined, such as, for example, if a stroke was drawn clockwise or counterclockwise.
We propose a 3D analysis by an opto-electronic system, in order to improve the graphology analysis for off-line signature verification.
I will bet my life savings that it cannot detect a fake signature with my signature. I sign my checks differently almost everytime and I dont have one set signature because I suck at writing script. Id like to see how it would fair then. I sometimes scribble my name, while other times ill write the letters out.
So what if I use something else like a gel pen? I do use those to sign check, you know.
The simple truth is that interstellar distances will not fit into the human imagination
- Douglas Adams
So obviously we all press down to the same intensity and the curves of our writing remain the same all the time! Screw signatures off soon it will all be done with biometrics. thumb print obtained & verified you are who you say you are.
3D Holograms Detect Fake Signatures
Several sources reported last week that a new technique that produces 3D holograms of handwriting could be used to detect fake signatures on checks, credit card receipts or other important handwritten documents. Here are pointers to Nature , Scientific American or BBC News Online . Instead of using 2D techniques to look at the sequence of pen strokes in a signature, this new method is based on 3D micro-profilometry which permits to translate the writing into an image showing dips and furrows of the sample so that anomalies can be detected. If you plan to imitate your spouse's signature, beware! Forensics have a new and very efficient tool. As an example, for the use of ballpoint pens on normal paper, the success rate was 100%.
Nature describes the problem and its solution.
Let's turn to BBC News for more details.
Here is a an example of "profilometric acquisition by means of conoscopic holography. These strokes were made by a BIC pen on common paper. The investigation area is about 5 mm × 5 mm. (a) 3D view of the strokes' profile. It is possible to note the regularity in the (S) line. (b) 3D view of the strokes' profile. The presence of bumps is evident. (c) 3D view with a mirror along the z-axis."
The research work has been published by the Journal of Optics A: Pure and Applied Optics in its Septemebr issue under the name "Superposed strokes analysis by conoscopic holography as an aid for a handwriting expert." Here are two links to the abstract and the full paper (free registration needed, valid for 30 days, PDF format, 6 pages, 320 KB). The above images come from this paper.
How is this technique working? Surprisingly well, according to Nature.
If you want to see the
This is amazing, although I'm surprised by the fact that it isn't already in use *today*! Detectives already have known that when you write on paper, it creates a depression in paper. If you're writing on a note pad, for instance, and after you write your address on the sheet above of a bank holdup note, just lightly rubbing a pencil against the hold up note will reveal the address, the one written on the sheet above the bank holdup note, all because pressure created an indentation in the paper. Plus I do believe people sometimes check when they get rewards, certificates, etc.. to see if the signature is really hand-made, or printed, just by feeling the back for an indentation suggesting that someone was writing on the paper.
Ah, you found me!
What about the electronic signature pads in use at many stores, well, everywhere? I used to work at a 7-11, and I can tell you firsthand that the resolution *SUCKS* on these things. Nevermind that they don't take pressure readings. When we first got them my manager signed it with her real name, then "Micky Mouse". I couldn't tell the difference. It'd still be useful for things traditionally signed on paper (insurance policies, etc..), but as far as debit and credit transactions the majority (and a growing number of) of transactions will be unverifiable by this method.
E pluribus unum
A 100% detection (at least in tests) of false signatures? Great!
How hard did they try to create a false signature?
And how often were legit signatures rejected? (I can create an algorithm that filters out 100% of the false signatures, guaranteed - it simply rejects all signatures it gets)
+++ MELON MELON MELON +++ Out of Cheese Error +++ redo from start +++
How do I use the DMCA against these bastards. My signature is my property, I don't want any reproductions of it stored in their computers - 3D or not.
Just how much closer are tablet PCs to true handwriting recognition? Authentity aside, handwriting appears to be a simple enough problem, but so much emphasis is placed on context. People can write according to lines on a page. A semicolon shouldn't be confused for an i.
Now if we can detect forgeries with science, surely the science can be programmed to decide whether I wrote the number 1 or the letter l or even know the letter t is not the letter f and the > is not 7.
Know your pads. One time pad: good for cryptography. Two timing pad: where to take your mistress.
Why does /. keep posting articles submitted by this guy? He has a shabby blog on radio.weblogs.com and does a poor job stealing other writers work; the site is a blatant commercial effort. Yet /. keeps putting Roland's stuff up and linking to it.
What's the deal? Is there some kind of commercial payola a la 1970's radio? Maybe the editor has a thing going with Roland, in a Clinton-McGreevy-esque way.
*Cringe* I didn't need any of those mental images.
and i can barely read my own signature!
seriously, i sign hundreds of documents every day. what happens when i can't replicate my own writing?
Supreme executive power derives from a mandate from the masses, not from some farcical aquatic ceremony.
I sign my wife's name all the time, and she sometimes signs mine. All perfectly legal, as long as neither one disputes it.
Simply claim you have power of attorney from your spouse. If, when asked, your spouse says "yes, I gave my permission", you're clear.
Of course, you better be DAMN SURE your spouse is going to back you up.
"yes, he did. Same way he gave me permission to sign his name on the check buying the mink coat..."
Learning HOW to think is more important than learning WHAT to think.
Ballpoint pens are probably the best choice for signing checks and other documents. The ink is hard to bleach and the ball crushes fibers in the paper, providing a record of pressure applied to the pen. Unfortunately, nobody bothers to look at signatures on most documents, like checks.
Mea navis aericumbens anguillis abundat
I'm not sure there's any point trying to prove a signature is authentic -> How many of us actually have formal signatures that we can replicate perfectly? I know sure as hell that my signature is different every time and, sequence of pen stroke or no, I'd fail a test for my own signature every time. I'd instead be working on a way to place an electronic signature - some kind of stamp or something, that acts like a GUID for a person how needs to sign something.
Would be 4d recognition. The 3 dimensions already listed in the article combined with time. Essentially, a device could record the 3 dimensions as you sign.
Liberty.
My signature can't be "read" (while it originally derived IHandwrittenName, it was significantly personalized and aesthetically simplified over time), and this has led to quite a few inane "Haha! THAT'S your signature?" comments.
A signature is just an individualized sequence of muscle movements that technically could be you writing an offensive remark. That's why there's normally a printed name aside it.
1. No one signs the same exact way twice. That's how some forgers get caught -- copying TOO exactly.
2. We have all developed habits. Although your signatures may look different from each other, the pressure patterns are usually identical. Forgery detectors use magnifying glasses to detect dicontinuities in the letters or words, indicating a lifting of the pen for a glance at the original being copied. Most people do not lift and replace the pen on the paper while signing their own name.
3. Some forgers use the trick of holding the signature being copied UPSIDE-DOWN so they can "draw it" instead of writing it. That way they avoid the traps of their own habits showing to an investigator. They are usually the good ones who escape being caught. This technique woiuld easily show that the signature was drawn upside down and last letter first, and they will be caught.
4. The harder the signature is to read, the EASIER it is to forge. My own signature is perfect Victorian calligraphy done with a chisel-point felt-tip or fountain pen. Let them copy THAT!
5. Most organizations never check signatures until there is a anamoly. By then, the pro forger is long gone.
6. Pro forgers will defeat these machines by practicing their marks' signatures until they are perfected.
As they always have.
If we have the technology to read signatures, why not just make a device to write signatures? Surely it can't be that hard?
We may not have the tech to exploit it in front of someone at the moment, but I can't imagine a laboratory-style exploit is far off.
This kind of revelation about how to protect against forgery ends up defining the victory conditions a lot better, giving attackers a clearer target.
Anyone want to take bets on how soon we'll see an article on an industrious group of technologists who modified a plotter to make automated signatures that cannot be detected as forgeries?
-- Fratz, human
The system is ancient but wide open to abuse. Several years ago a woman returned from holidays to discover that she had been married to her workmate. He had simply obtained all the paperwork, stamped it with his own seal and then having taken hers from her drawer, stamped it with her seal as well. The marriage was anulled, but the point is the personal seal is a little dangerous in my opinion.
Anyway, as a result, very few people I know in Japan has what I would call a signature, that is something that you write almost the same way each time. My own signature varies each time I write it an in fact has shrunk over the years, but always contains elements that appear to be hard wired into my hand and brain now. Even if I use a different grip, or even the wrong hand, the pattern is similar (thought obviously different). When my students try to sign something, they usually very neatly write their name in English.
Most modern handwriting recognition techniques, such as on my Tablet PC, rely on knowing the order and direction of writing strokes to improve their accuracy. It looks like the techniques described by TFA's sources would provide similar information and might enable machines to finally transcribe handwritten papers reliably.
My new
If THAT is true, the children will not understand anyway :-)
I suffer from several severe forms of anxiety, performance anxiety being one of them. This includes the simple task of signing your name infront of someone in a bank or elsewhere. It is really aweful when you can't even feel comfortable signing your own name when in the presence of someone else.
My signature changes regularly as I seem to mess up or I can't seem to do it the same every time, it really is fustrating.
"Software is like sex... it's better when it's free"
As for myself, I share the sentiments of the original poster. My writing is so messy and signature so random, I don't think a system like that would work so hot for me. I guess I do have a few consistencies, like the elongated loops on my capital A and lower D's, the goofy squiggle that's supposed to be a capital W, and how I put a double stroke through my name when I dot the i and cross the capital A all in one fluid motion (thought it looked cool when I was little... now its just a part of my signature. Its funny when people asking me "why'd you cross your signature out? You just signed it!"
Oh well, at least with my mother working at my bank (branch manager no less), no one questions me on stuff like that.
Teller: "What the hell? Is that supposed to be your signature"
Me: "FIRED!!"
My signature has always been inconsistent. Hell, letters start turning into odd scribbles when I'm in a hurry. I've seen my 'z' turn into an atrocious, indiscernible jumble of curves. It's close enough for someone at the bank to glance at and say it's fine, but if someone was actually going to look at it in detail they'll probably think it's forged.
You can get much more detailed information about how a pen is held, about the timing and order of strokes, and how much pressure is excerted with a modern computer tablet. Even if you have all that information, you don't get anywhere near "100% accuracy" for signature verification. Since the data they work with contains less information, we can pretty much conclude that it must be their experiments that are poorly done, not that they have hit on some amazing new technique.
That isn't to say that the technique is completely useless. But it won't solve the problem of document forgeries.
One question one needs to ask, however, is whether the authors have any connection to the maker of conoscopic holography equipment...
I just RTFA (sorry) and had a thought - this technique could be adapted to a device that could read the information from an LP or an old wax cylinder without touching it.
This could make a recordings museum caretaker very happy. He could hear the recordings that are too fragile to play.
If there was a contact-free record player, I wouldn't feel like I should sell all my LPs.
I remember a story in Analog a few years ago about a man with the only recording of his father's voice on an old lacquer disk which had unfortunately broken. He ends up being able to listen to it due to a tech not unlike this.
A good story. Damn, I miss having the time to read those every month.
An affordable application of this for repairing a damaged record (not just applying filters) would certainly end up on my dock.
Where's Robin Hood? We could kinda really use him now.
They're "AutoPen."
The first time I went to the US, I was astounded to find store clerks swipe my card and then hand it right back to me before I sign the chitty!
My wife (American, now living in the UK) gets peeved that she can't go shopping with my credit card, because here, shop assistants are trained to CHECK the signature before letting you buy something......
What a novel idea.
Just finished a part time job at a bank where I learned a frightening amount about check fraud. Such technology will do NOTHING to stop professional check cons... Most fraud goes in the check-by-mail or ATM system with enough co-deposits to ensure that the check is never looked at once by a human being, at least not for authenticity or the like. The nice little numbers on the bottom of the check do the trick and the magnetic reader takes it from there. When a person does "proof" or look at the check they will focus soley on the amount of the check and the account number... They have WAY too many items to look at such silly stuff like.... name, pay to the order of or... signature. The looming Check 21 Act aims to stamp out some fraud, but won't do much most likely. How much money would it cost banks as a whole to employ enough people to inspect each of the millions and millions of checks each day? Tellers see only a portion of a bank's total check activity. With check fraud at some $500 million a year I guess the banking industry decided it's cheaper to live with than to fully combat. Doubt some new 3D sig will do anything at all to help...