Life Casting for Nefarious Purposes
-
Not really what I'm thinking about. I mean how much detail can it capture, not how long can it retain the detail. If I capture an image at 320x200 pixels and store it on optical media, it'll keep for a very long time indeed. But the quality is still pretty low. What kind of resolution will the best masks have? Has any actual research been done on this?
Don't let my name fool you. That's my job.
You misunderstood. My comment is on how long the human face remains exactly the same, so how detailed can your recognition be as it needs to recognize the real face over and over again. BTW: Vista is great, unfortunately I don't have any anecdotes to illustrate that. :)
Luc Pattyn [Forum Guidelines] [My Articles]
- before you ask a question here, search CodeProject, then Google - the quality and detail of your question reflects on the effectiveness of the help you are likely to get - use the code block button (PRE tags) to preserve formatting when showing multi-line code snippets
-
You misunderstood. My comment is on how long the human face remains exactly the same, so how detailed can your recognition be as it needs to recognize the real face over and over again. BTW: Vista is great, unfortunately I don't have any anecdotes to illustrate that. :)
Luc Pattyn [Forum Guidelines] [My Articles]
- before you ask a question here, search CodeProject, then Google - the quality and detail of your question reflects on the effectiveness of the help you are likely to get - use the code block button (PRE tags) to preserve formatting when showing multi-line code snippets
No, I got that. It just isn't that important. If a user has an accident or starts getting older or something, and they generate a lot of false negatives, they can just re-enroll their "new" face. Problem solved. But that doesn't help me find out if a face can be faked.
Don't let my name fool you. That's my job.
-
No, I got that. It just isn't that important. If a user has an accident or starts getting older or something, and they generate a lot of false negatives, they can just re-enroll their "new" face. Problem solved. But that doesn't help me find out if a face can be faked.
Don't let my name fool you. That's my job.
But it presumably copes with, for example, bloodshot eyes, not having shaved for a day or two, different earrings (or other piercings) and different makeup? If this is the case, then what is the level of detail it is picking up on? How many data points? If it's using infra-red I would have thought you would need to wear the mask (or heat it to body temp) and that might give the recognition software a good opportunity to spot a dud? I read about a bank (i think it was a bank) using facial recognition software, but in tandem with security cards - so it checked if the face belonged to the person whose card it was. Obviously Charlies' angels would have to steal/copy the card too.
___________________________________________ .\\axxx (That's an 'M')
-
The physical security company I work for has a question about the effectiveness of 3D face recognition technology. It works by scanning the face with special cameras to measure not just height and width but also depth. It creates a database of points to describe each face it reads, and compares that with stored templates to find a reasonable match. It's pretty good at saying Bob's face is not Fred's face, but it's not so good at saying whether Bob is a mannequin or a real human. While testing, we successfully enrolled both a mannequin and a face mask, and then went on to have those two items validate, thus opening the door of our high security room (well, if we had actually attached it to such a door, which we didn't). This naturally brought up the following spy scenario: Charlie's Angels slip Mr. Big some rohypnol, then make a mold of his face while he is unconscious. He wakes up a few hours later thinking he simply passed out drunk, while they create a silicone face mask and gain access to all his company secrets. Sounds silly, but isn't this possible? Google doesn't want to tell me with any detail. I am hoping someone out there has information about the accuracy of such a face mold. Can you just get their major skin features, or can you even approximate large pores? How likely is it that you could take a cast while they were unconscious? Would being unconscious make your face deform slightly? As you might expect, the face reader manufacturer is not taking this inquiry seriously. But is that because it sounds so implausible, or because they know their product could be fooled?
Don't let my name fool you. That's my job.
I prefer biometric systems that rely on thumbprints. It's too hard to cut off an employee's face.
"A Journey of a Thousand Rest Stops Begins with a Single Movement"
-
The physical security company I work for has a question about the effectiveness of 3D face recognition technology. It works by scanning the face with special cameras to measure not just height and width but also depth. It creates a database of points to describe each face it reads, and compares that with stored templates to find a reasonable match. It's pretty good at saying Bob's face is not Fred's face, but it's not so good at saying whether Bob is a mannequin or a real human. While testing, we successfully enrolled both a mannequin and a face mask, and then went on to have those two items validate, thus opening the door of our high security room (well, if we had actually attached it to such a door, which we didn't). This naturally brought up the following spy scenario: Charlie's Angels slip Mr. Big some rohypnol, then make a mold of his face while he is unconscious. He wakes up a few hours later thinking he simply passed out drunk, while they create a silicone face mask and gain access to all his company secrets. Sounds silly, but isn't this possible? Google doesn't want to tell me with any detail. I am hoping someone out there has information about the accuracy of such a face mold. Can you just get their major skin features, or can you even approximate large pores? How likely is it that you could take a cast while they were unconscious? Would being unconscious make your face deform slightly? As you might expect, the face reader manufacturer is not taking this inquiry seriously. But is that because it sounds so implausible, or because they know their product could be fooled?
Don't let my name fool you. That's my job.
-
But it presumably copes with, for example, bloodshot eyes, not having shaved for a day or two, different earrings (or other piercings) and different makeup? If this is the case, then what is the level of detail it is picking up on? How many data points? If it's using infra-red I would have thought you would need to wear the mask (or heat it to body temp) and that might give the recognition software a good opportunity to spot a dud? I read about a bank (i think it was a bank) using facial recognition software, but in tandem with security cards - so it checked if the face belonged to the person whose card it was. Obviously Charlies' angels would have to steal/copy the card too.
___________________________________________ .\\axxx (That's an 'M')
It determines spatial position based on a very specific area of the face. Seems to be the eyes, nose, and lips. Ears are definitely outside that range, and I'm not too sure about the chin. The marketers claim up to 40k data points, but that's not what it typically stores. There are about 20 main data points as mentioned, and then points surrounding those areas are included. It doesn't really have any kind of "liveness" detection. It's not checking heat signatures at all, AFAIK. It's only using surface detection, so bloodshot eyes (or even a sudden tan) don't seem to matter, but glasses can interfere. Most companies I've seen _do_ use card + face reader when they want higher security than either alone. But if the face reader is easily fooled, that kind of defeats the purpose.
Don't let my name fool you. That's my job.
-
But which one? The movies or the original series or the remake series? And what about Man from UNCLE? And the Avengers? I'm pretty sure there was at least one Bond movie that had this, and I know the X-Men had something similar (although that is pure fantasy).
Don't let my name fool you. That's my job.
-
The physical security company I work for has a question about the effectiveness of 3D face recognition technology. It works by scanning the face with special cameras to measure not just height and width but also depth. It creates a database of points to describe each face it reads, and compares that with stored templates to find a reasonable match. It's pretty good at saying Bob's face is not Fred's face, but it's not so good at saying whether Bob is a mannequin or a real human. While testing, we successfully enrolled both a mannequin and a face mask, and then went on to have those two items validate, thus opening the door of our high security room (well, if we had actually attached it to such a door, which we didn't). This naturally brought up the following spy scenario: Charlie's Angels slip Mr. Big some rohypnol, then make a mold of his face while he is unconscious. He wakes up a few hours later thinking he simply passed out drunk, while they create a silicone face mask and gain access to all his company secrets. Sounds silly, but isn't this possible? Google doesn't want to tell me with any detail. I am hoping someone out there has information about the accuracy of such a face mold. Can you just get their major skin features, or can you even approximate large pores? How likely is it that you could take a cast while they were unconscious? Would being unconscious make your face deform slightly? As you might expect, the face reader manufacturer is not taking this inquiry seriously. But is that because it sounds so implausible, or because they know their product could be fooled?
Don't let my name fool you. That's my job.
These days you wouldn't need to create the mould on someone's face. I've seen a tool that is used in the car industry for copying the surface of clay models where light beams move across the surface and the detectors create a full 3d surface model in the computer. I guess the smaller the beam and the shorter the wavelength the more accurate this system would be. I don't know how small one of these machines could be or how much they might cost but the technology is definitely out there. From there it's a relatively simple task with a milling machine and a block of something solid to create a mould for your mask.
-
Actually, the mask wouldn't need to be worn at all. Just presented to the camera. What I really want is to find out what the granularity is, since the manufacturers talk about "micrometer precision". We've gotten some flaky claims from some of the salespeople that were later debunked by their technical people, so we don't trust non-specific claims anymore. For example, we were told both that it measures down to bone structure and that it measures skin softness. Both claims were outright false, since it only measures the infrared light reflected off the skin. But their technical people are not taking this spy scenario seriously, and I want to know if I can show them real facts or if I'm worried over nothing. Because we ultimately have to explain this to _our_ customers, who are starting to mistrust us for repeating those damn marketing claims.
Don't let my name fool you. That's my job.
-
I prefer biometric systems that rely on thumbprints. It's too hard to cut off an employee's face.
"A Journey of a Thousand Rest Stops Begins with a Single Movement"
You know, you could also log on to the DC, navigate the Active Directory, select the user in question, open his settings, and remove his membership in the "UsersWithAccessToFloorSeven" group...
Don't attribute to stupidity what can be equally well explained by buerocracy.
My latest article | Linkify!| FoldWithUs! | sighist -
I prefer biometric systems that rely on thumbprints. It's too hard to cut off an employee's face.
"A Journey of a Thousand Rest Stops Begins with a Single Movement"
Roger Wright wrote:
I prefer biometric systems that rely on thumbprints. It's too hard to cut off an employee's face.
... but a head is easy enough! Incidentally, when I was a student back in the 80s - we had a demonstration of a neural network based facial recognition systen (if I remember correctly it was called Wizard). This did not attempt to match small specific features but, but due to the neural net, more of an overall impression of the face. It seemed very effective and had no trouble in recognizing people who adopted a false moustache or glasses and I seem to remember (could be wrong) that it was pretty effective at distinguishing twins. Input was a simple B/W video camera. These days, the rack of hardware could be replicated in software and the technique could, I would have thought be easily replicated on a PC. Must dig out those ancient lecture notes on NN ... Rich
-
but I'll answer yours and hope you'll consider returning the favor. The most obvious difference from a card is that you never accidentally leave your face at home or drop it when getting money out of your wallet. Thus, there is less chance someone can steal it. Normally, combined with the uniqueness factor, that would make face readers more secure than cards. It definitely makes them more convenient to the user. --- We can deal with the other things you mentioned. Those are just "common sense" in the security industry. The problem is that I've outlined a possible theft scenario and I cannot find out how to evaluate the relative risk. I don't have the expertise with life casting (nor the funds) to conduct valid experiments.
Don't let my name fool you. That's my job.
Naruki wrote:
but I'll answer yours and hope you'll consider returning the favor.
I'll try, but I am not an expert in that area.
Naruki wrote:
The problem is that I've outlined a possible theft scenario and I cannot find out how to evaluate the relative risk. I don't have the expertise with life casting (nor the funds) to conduct valid experiments.
I don't know how to evaluate the risk factor either. But, I tend to think, it is very low risk to the point of I decide not to handle this case. Because, unlike a theft where the objective is to gain an access to an object and run away, in this case, there is some work to be done. Also, there would be some complications trying to cast the subject. Unlike real person who is in control of him/herself, the knocked down person could move, wake up or perform sudden movement of the face which might hinder with the casting process. Another possible scenario, which might be higher risk is to force the subject at gun point to follow your order. In this case his/her face will be used to get an access to the secure area. just my 1.99 cents
Yusuf Oh didn't you notice, analogous to square roots, they recently introduced rectangular, circular, and diamond roots to determine the size of the corresponding shapes when given the area. Luc Pattyn[^]
-
I prefer biometric systems that rely on thumbprints. It's too hard to cut off an employee's face.
"A Journey of a Thousand Rest Stops Begins with a Single Movement"
Hannibal Lecter used that very technique to good effect.[^]
Software Zen:
delete this;
Fold With Us![^] -
The physical security company I work for has a question about the effectiveness of 3D face recognition technology. It works by scanning the face with special cameras to measure not just height and width but also depth. It creates a database of points to describe each face it reads, and compares that with stored templates to find a reasonable match. It's pretty good at saying Bob's face is not Fred's face, but it's not so good at saying whether Bob is a mannequin or a real human. While testing, we successfully enrolled both a mannequin and a face mask, and then went on to have those two items validate, thus opening the door of our high security room (well, if we had actually attached it to such a door, which we didn't). This naturally brought up the following spy scenario: Charlie's Angels slip Mr. Big some rohypnol, then make a mold of his face while he is unconscious. He wakes up a few hours later thinking he simply passed out drunk, while they create a silicone face mask and gain access to all his company secrets. Sounds silly, but isn't this possible? Google doesn't want to tell me with any detail. I am hoping someone out there has information about the accuracy of such a face mold. Can you just get their major skin features, or can you even approximate large pores? How likely is it that you could take a cast while they were unconscious? Would being unconscious make your face deform slightly? As you might expect, the face reader manufacturer is not taking this inquiry seriously. But is that because it sounds so implausible, or because they know their product could be fooled?
Don't let my name fool you. That's my job.
As someone else said, why bother with drugging the person in question? A mask would take a while to set, so you'd have to keep the person out for a while. A pair of decent camera's at known positions can do a 3d scan, rapid prototype a head with a hole in the top, pour in hot coffee to fool the infra red. The best advantage of this is you can scan a lot of people walking by the cameras, then pick and chose your victim. Even the best facial system out there isn't worth much more than protecting a laptop. If you want high security biometrics, you'd have to go with retina scans. They're more difficult to steal and replicate. Couple that with personal pin numbers and you should be at the top of automated security.
The true man wants two things: danger and play. For that reason he wants woman, as the most dangerous plaything.