The media (because they don't understand security), and certain biometric firm executives and salespeople promoto this view - it is a quite horrid view as well. Every engineer and developer I met in thefield (and I met quite a few) thinks it is disgusting - but the media always wants to interview the executive or marketing guy in the suit - not the people who know how the things work and should be used.
Uhm, I am confused, did you read my post and it's description about how multiple layers of security are really needed? - Even Schneier agrees with this principle - it is the use of a biometric instead of a password, which I explicitly mention is not smart, that he is against, and that he thinks adds a false sense of security.
You seem to assume in this post that a biomtric be used on its own, and then discuss failure rates that are low (for authentication), but about right normally for identification (except for facial recognition systems, where it is lower).
You then go and claim that 100 people would be 'accused' of being terrorists, when my text clearly discusses how that is the Wrong way to use such a system - that each claim of such a system needs to be looked at closer, but the system is not to be trusted to be accurate - hence using another camera to zoom in on possible match's face, and get a better reading. The vast majority of a system's failure rate is due to the quality of samples taken.
It is the improper use of such systems, as I described, that leads to being being falsely accused - and from that, we get horrible impressions of the systems themselves, not the (untrained, illinformed) people who run such systems. It is very easy to a security cheif who is getting in trouble because his team is falsely accuwsing people to point to his new facial scannign system and go 'it sucks!' and quit using it, rather than to look into why 'it' isn't working properly and find out that his people don't know how to utilize such a system.
Do you honestly think A bank would use such a system? In my post, I was speaking of reputable systems, not the crappy 200$ versions bought for price over features.
As for subscribing, I already do - and having been a developer in the industry for years, I have been reading up on security, as well as holes in biometric systems, for quite some time. But just because device A from a year ago had issues, does not mean new devices will - since any manufacturer who wants to survive will avoid the mistakes of the past.
As for your assertion on facial recognition problems - one system common in use != 'Most facial recognition systems CAN be fooled by holding up a picture'. Usualy, such problems are incurred via the source of the image, when using facial recognition systems that is. Visionics 'FaceIT' for example, becomes less and less secure with cheap cameras. The cost involved in improving such systems is actually very low - the difference between usign a webcam as the image source and using a 2+ megapixel camera. That is 100$ - that does not limit the usage in military bases by any means. Lighting and other proper setup features are usualy just as important for a useful system.
As for authentication versus identification issues, especially in airports, see my post here: http://slashdot.org/comments.pl?sid=88410&c id=7655 259
your point about a false sense of security, and the need for multiple layers of security in an authentication scheme is correct, but so much of the rest of your post is incorrect, so I feel the need to interject.
- a stolen biometric isn't useful except agaisnt the same sort of scanning system - as in, the same manufacturer. No standard data format exists. - the pin example is a bad one - the theif needs your card as well (as it is the other layer of security in the system). Anyone who gets the biometric data from the thief will have a hard time using it if they also need the new, shiny, replaced bankcard. - most biometric systems can tell the difference between dead and living tissue - although this might not stop an ignorant criminal in the first couple of years, it would become commong knowledge that the cut-out-the-eye trick doesn't work once some people ar behind bars.
I think you forget that in any case of security, another weak leak is relying upon any single means of identification/authentication.
The example you gve of the database being compromised is a very horrific one - but the same sort of problem exists for passwords.
I should probably mention that, jsut as passwords are usualy stored as one-way-hashes of the actual plaintext, so biometrics are stored as one-way processings of data taken from a scan, so in either case, the original data (the plantext/scan) isn't being stolen, its representation is being replaced.
Your scenario would be deadly for any authentication scheme, even one that used multiple types of authentication (as all truely secure systems do). Normally, this would be two or more of the following types of items: - something kept (a security card, id card) - something known (a password) - soemthign intrinsic (a biometric, a dna scan)
But all three fail if the root system is compromised that contains the information they are being compared against...
So, your scenario is a problem, but not just for biometrics.
Ok, first, your post seemed directed at me, but was a reply to the parent.
Secondly, how can you claim no ROI for such a rollout? If the return on investment is enhanced security, and if such a system were to provide you with better results than just having extra officers standing around with sheets containing photos of known bad guys, then the ROI is positive.
However, the bad guys must be identified Before they came into an airprort or some other area with such a system - otherwise the system isuseless, as would be name-lists and security guards and wanted posters.
Pre-9/11, most of the hijackers were not on the rader of law enforcement. That sort of failure can ruin investments in all sorts of security technologies - not just biometric ones.
- biometric data is not stored as a simple image. It's not stored as a compressed image, or a md5 of the image. It is most often stored as a series one-way-hash values, each of which is derived from some characteristic inherint in the scan. Someone could steal this data, but creating the original image is near impossible, like breaking a 100 kilobyte rsa key. - biometric data is stored in a different format by every manufacturer. There is no standard - heck, they can barely get a standard API for how to interface with the hardware and drivers (www.bioapi.org), let alone agree on a standard format. Thus, if visa were to start using scanners, and your fingerprint scan were stolen, only visa systems would be affected. - most authentication systems (other than the implied example of logging onto a computer) use multiple pieces of information, usualy two or more of the following type:
- something remembered ( a password or pin)
- something kept (a security card, a credit card)
- somethign intrinsic (a biometric)
Now, how useful is that fingerprint scan if the visa card it's associated with is not in the theif's hands? How useful is it if you cancel your card and get a new one?
- if someone did manage to steal an image of your fingerprint or retina, it won't do much good: systems these days are able to tell the difference between a dead/living finger, a photo, and even a plastic mold (many systems look for temperature of what is scanned, and can even look for capilary blood flow).
- if someone gets access to a computer system where they can use the information stolen and bypass the scanning device, well, you have much bigger problems: such a breakin would probably compromise things to the point where they can simulate a positive authentication from the driver/hardware, for any user.
- (this one only applies to fingerprints): you have ten fingers, use a different one. For eyes, switch eyes.
Having said all of that, please realize that biometrics are intended to enhance security by adding another layer to the authentication systems in place, not to replace them. A bankcard+pin+fingerprint is more secure than a bankcard+pin.
Anytime you hear/read the mass media promoting the death fo passwords via biometrics, realize that either A) the reporter doesn't get it or B) they have talked to a marketing person at one of the manufacturers who is (most likely in my experience) pandering to the media in an attempt to grow the market and get sales, despite the falsehoods involved.
By the same token, anyone who tells you a password by itself is secure, is also wrong.
AS someone who worked at a company that tried, pre-bapi, to write our own abstraction layer over multiple vendors' drivers, I feel your pain - our entire biz model was to write our own applications in such a way as to be used with different devices, and provide our sdk for other developers to do the same. The more applications, the more demand there would be for the devices - just like video cards with 3d acceleration and games.
What we found was telling in several ways: - Device manufacturers saw their SDKs as profit center, and did not want to give them away - even though it would promote device sales. - The engineers at many firms were very, very intelligent about biometrics, but not about software - they would buy driver writing kits for windows and go from there. Sometimes they were very smart electrical engineers as well, designing not only the authentication schemes and putting cameras into little plastic boxes, but designing their own special circuitry to be used on a pci card to provide a high bandwidth interface to thier device, etc. But once again, after going to all that trouble, the firms used driver writing kits.
Of course, what you seem to be missing, is that these vendors have (what appear to them) to be good reason to Not provide open source drivers - or binary drivers for an OS that is run by users who will just try and reverse engineer them.
- often, the actual authentication mechnisms were performed in sofware, with the device just gathering data. Thus, by providing drivers they would open up the core of their product and simply become hardware makers. - there is no economic incentive - not only does the above point illustrate this, but the demand hasn't been there from the people who actually buy the systems.
Having said all that, note that some closed-box systems that do not require an external computer are rumored to run embedded linux variants...
I think you need to look into security principles. As you say, a lone password is easy to compromise, so is a lone biomtric. However, any truely secure system needs to use multiple forms of identification - preferably two or more of the following: - something intrinsic (a biometric, dna scan, etc) - somethign known (a password) - somethign kept (a security card)
By having more than one step involved, the system is much more secure than any individual part. Somesteals your backcard - but do they have your pin? Or, someone sees your pin - but do they have your card or account number? PINs are actually very simple and easy to break (thoeretically), but are pains to break in reality because of the Other required piece of the puzzle, the bankcard, and how false authentications lead to the removal of the card (most ATMs shred your card after a few false PINs are entered).
similarly: Just because someone steals your face, how will they get ahold of your new bankcard?
After that fact comes the fact that most biometrics are hard to fake - fingerprint scanners these days can be made smart enough to check the temperature of the item placed on them - and some are even smart enough to look for normal temperature differences and gradients within the skin surface, and refuse authentication to 'fingers' that are too regularly or irregularly warm. Some very high end systems look for capilary blood flow... Most facial systems are smart enough to refuse a photo held up of your face, and carrying around a stiff 3d mask of someone's face is kind of obvious.
Also, the fact that every type of scanning device on the market practially has a different data format for the biometric data (which is all one-way, you can get the data from a fingerprint, but not the other way around), and spoofing the data becomes more restrictive - a spoof of, say, visa's system wouldn't work against mastercard's (unless they were using the same equipment).
Having said all that, I'd still like it to be pin+card+face/fingerprint rather than card+biomtric. Biometrics should be used to Enhance security, not replace known or kept-item security methods.
Facial recognition is only 1 of the technologies involved in biometrics... To claim that the whole industry has failed to grow because one Type of biometric does not function well is untrue.
Besides that, your numbers are wrong... facial recognition systems can actually have failure rates higher than that under less than ideal ircumstances, and when put into use as identification, not verification systems.
First, definitions, for those who didn't read the article:
Identification: determin from a scan who someone is, searching over a list of possibilities.
Authentication: determin with reasonable confidence that the user is who they claim they are.
Authentication is much much easier to get right, since you can always ask for a rescan if you are unsure. Authentication systems are designed so that the device (hardware and software) return a confidence level - sometimes a percentage. It is up to the application developer to determin just how high a confidence level you want. If you set it too low, people with similar faces might be abel to authenticate for each other - borthers for example. If set to high, then slight (natural) variations in a person's face can cause rejections. Generaly, you must strike a balance between false positives and rejections. Such a compromise is acceptable, if you have other security measures in place (see note at end of post).
Identification is much, much harder. First of all, it is very cpu intensive - one can model identification as a low-confidence-level authentication against every listed person in the database. If you have 40,000 people in the database, this can take awhile. Hashing doesn't help much, and is illadvised, since we are looking for a close match, not an exact. Biometric data isn't the kind where you can take the first 5 bytes and dump into hash buckets either - but I digress. So, how do you speed it up? You reduce the dataset by reducing the detail in the data you store for each person.
Then you run into the problems with how these systems have been rolled out - using low resolution security cameras is not a good way to get an accurate scan of a person's face - especially when the people being scanned a re small enough (in relation to the scene) to be only 10s of pixels wide.
So, now we know the technical difficulties - but why the bum rap, and why would a police force choose to roll something like this out anyway? This is several fold, but the main thing it comes down to is misconceptions about what these systems are doing, and badly written systems. Due to the limitations mentioned above, these systems can only provide possible matches, like 'Person X is a 20% match against Osama Bin Laden'. the system isn't claiming that the person IS Osama, only that the face appears somewhat similar. As such, the system is supposed to be used as a guide - if it picks someone out, that person deserves more attention - that attention could be a remote-controled security cam singling them oout for a better scan, or for officers in the area to walk over for a better look. Unfortunatly, just because that is how the system is supposed to work does not mean it is used that way - all too often these are rolled out as a way to 'increase security while retaining a minimal police/secuity force'. You get officers who think of a potential match as a authentication, and they send officers running down at high speed only to find it's not Osama... The next potential match they are more hesitent about, and so on, until they mistrust the system completely. Is the system doing anything wrong? No, its that the users don't understand what it is doing. Better training would help, but so would the people making the purchasing descisions understanding the technology, and staffing accordingly.
In the sort of rollouts described above, facial recognition has a success rate of less than 30%, much lowe r than what you describe. With rates that low, people complain, and stories get published. Used properly, the data these sy
The economist article fails to mention the other major reason these systems have not taken off - comparability.
Or, I should say, the Lack of it.
Each fingerprint device on the market uses its own format for storing it's data - making each device incompatible. At first, this would seem to be an easily surmountable problem - but then you must realize that until recently, Every device on the market had its own API for development.
Let me give you an example to illustrate this issue: company X has 2000 employees, and it goes to look at biometric systems - they are either faced with the choice of paying for very expensive equipment from 'long time players' in the industry - who would be around in 2-5 years when the devices start failing due to wear and tear - or choose from some of the 'upstarts', and risk being out in the cold if the company they choose isn't around in several years. a hardware switch down the line not only would incur the cost of re scanning everyone, but the application itself would need to be modified to work with the API for the new device.
Enter the BioAPI (www.bioapi.org) - which proposed a standard api - now widely adopted. You may notice that the Bioapi page mentions it was founded in 1998. It has taken several years for this standard to come to the foreground and there are still roadblocks - not all manufacturers participate freely. As an example: one rather large manufacturer, Identix (www.identix.com) seems to have been stonewalling for years. Why would a manufacturer do such a thing against what is good for the industry? Because they were leading the industry. When you have all of the high end government contracts coming your way, a standard the opens the doors for the little guy is a Bad Thing for your business - or so they thought. Take a look at the members list on the bioapi site - identix is listed - then take a look at the supported devices list... not a single identix product.
In 1999 I witnessed this stonewalling firsthand at a meeting in washinton DC. This meeting had manufacturers and interested parties from all over the globe in attendance, including representatives from the US military. The whole agenda for the meeting was how to promote/define standards so that the industry could grow. I had the unfortunate luck to be seated next to the Identix representative. He had apparently flown in just so he could stonewall - every opportunity he got, he grabbed the microphone and ranted about how we should let the free market dictate standards - that they would come about naturally in the free market (he loved the term free market). Meanwhile the rest of the group was discussing issues about how to resolve device inter operability - even so far as to discuss how data could be shared between devices. No concrete decisions were made at the meeting, but it did get people talking.
Anyway, my whole point is, one of the major reasons the biometric security industry hasn't grown (as fast as has been predicted for the past 8 years) is because without standards no one wanted to invest in writing applications. It was just too risky.
Note: I am flipping a coin as to wether to post this anonymously or not, since Identix could decide to try and silence this sort of talk...
I know that the new verizon specials are much better than the old plans - my 75$/month plan is now a 60$/month special.
The special is so good, I'm moving my other phones to it - previously that had 50$/month plans plus a 10$ feature-add (mobile-to-mobile minutes). The 60$ plan includes the mobile-to-mobile minutes and 400 more peak minutes than the old 50+10 I was doing.
So, I'll save 15$/month, and get 400 extra peak minutes on two of my phones. I'd hoped for even bigger savings, but 180$/year with more minutes isn't too shabby.
Now, if only I got good reception in all of my house - I'm apparently located at the edge of two cells...
Are you using Verizon? I know verizon lets you switch plans in mid-stream, as long as you get a new term on the selected plan that is longer than what remains on your current agreement. IE, if you have 14 months left, you have to get a new 2 year agreement.
I've never had a problem switching to new promotional plans since I got my phones (I have 3 phones in my name for myself, my wife, and the business - no land lines). This may just be a Verizon thing though.
Ah, but that is the problem itself - the poster I was replying to did not state 'there are problems with Darwins theory of natural selection' - he stated that there are challenges to evolution - which, although described by Darwin's theories, is not exclusively Darwin's. I know, I know, it comes down to semantics and what did the poster actually mean, but thats why I commented in the first place: because his phrasing would imply to the casual reader that evolution itself was in question - when in fact, it is only theories about how evolution works that are being questioned. Am I making sense? Maybe I was seeing a mountain where there was a molehill...
What you describe exists, but alas, it runs windows...
And is expensive for basicly a suped-up pda with a terminal services clients.
Look here: http://www.viewsonic.com/products/smartdisplays.ht m
If they were not quite as expensive, I'd buy one in a heartbeat.
Note: They use 802.11b to communicate with the host machine, and I do not know if they will work remotely very well.
If you want to save money, look on ebay for the predecessor to the, the viewsonic 'superPDA' - its basicaly the model 100 (see the business version) airpanel, but with a pda version of CE used instead of a stripepd down version for 'smart displays'.
Oh, and as for a slideshow while they recharge... uhm, I guess a screensaver?
Even now, there's substantial logical and statistical problems with the "proofs" of Evolution.
Are you refering to darwinian ideals of evolution, or the concept as a whole?
True original darwinism as the sole motivator for the changes in species over time is being challenged, but the concept as a whole - that life came from very simple beginings and has changed/adapted over time is not. The mechanisms involved are what are being challenged - such as the idea that small changes in genotype over time that favor the survival of a particular subset of a species lead to massive changes in the long-view. Fossils for the 'in-between' variants are not being found, hence it is becoming more widly accepted that large leaps are made, and that such large leaps could actualy be triggered by environmental pressure.
However, these new mechanisms being discused and discovered are just that - mechnisms. Evolution as Darwin envisioned it may be being disproven, but the idea that life evolves over time is not.
If, instead of refering to darwinian evolution, you are refering to evolution as a whole - then you are seriously mistaken. There is no creationist or other theory of life that is being pushed ahead of evolution by scientists. The logical and statistical problems you mention are about the problems with darwinian evolution and its mechanisms.
The bug you describe with 1.4.x and swing components is a known problem with the 1.4.1 series - though it seems to be fixed for me in the latest 1.4.2_02 release, and has been fixed since the 1.4.2 initial release.
The workaround is to disable direct3d and directdraw when using swing - they apparently do some flaky calls to directX libraries in order to speed up the gui drawing, and the ati drivers _hate_ it.
I'd refer you to the bug report on sun's site, but it requires registration, so I'll just cut-n-paste the workarounds as found on a non-sun site:
" Upgrade to Java 1.4.2
-or-
Set the Java command-line flag -Dsun.java2d.noddraw=true to disable these direct draw optimizations in Java 1.4.1.
-or-
Switch back to JVM version 1.4.0 which does not have the DirectDraw optimizations which trigger this problem in the ATI driver. This is a pretty undesirable solution, as many bugs were fixed in 1.4.1 and simple things like TreeTable fail to work in 1.4.0. Be sure to use Java Web Start v1.0.1. Web Start v1.2 is only installed with JRE 1.4.1.
-or-
Goto "Display properties -> Settings -> Advanced -> Troubleshoot" and drag the Acceleration slider all the way to the left to turn off hardware acceleration. (this will reduce but not eliminate the lockups) "
How is the parent, an obvious Joke if you follow the link, flamebait? if you don't like the humor, thats fine, but flamebait is a post designed to instigate an argument - I don't think anyone with 2 brain cells would say a silly joke is flamebait...
Thought Equity gets agreement from any and all parties to get exclusive worldwide distribution rights, Schaff said. Thought Equity also doesn't use ads featuring actors who belong to the Screen Actors Guild.
I am the troll? I am trolling? I made a joke, expanding upon someone elses joke. You didn't take it as a joke, and you went off, YELLING in your response... but I am a troll? I don't think you know the term properly.
Really, I mean, these links are all mostly off topic, and none of them address the issue directly of wether OS/2 ver 3 became NT. Your statements to the contrary in your post, and your (once again) yelling, do not disprove the basic fact that the two _are_ related and came from the same original project.
I mean, come on, does this link: http://tunes.org/~unios/oskernels.html
Has anything to do with the conversation other than to prove you would rather read a single paragraph description of what a microkernel is rather than read the actual source data? The books I provided links to on amazon are all college (and in some cases, graduate course) textbooks - I'm pretty sure they define a microkernel a little bit more in depth then what this link does.
Or did you think this link would 'teach me' about NT? This link provides no comparison between the NT kernel and the OS/2 kernel, and thus is useless in this discussion other than to be redundant.
http://pages.prodigy.net/michaln/history/os220/
This, once again, has nothing to do with the conversation - which is about OS/2 ver 3 that was underdevelopment when the MS-IBM split happened. Let me sumerize this page for you: OS/2 2.0 was 32 bit, but didn't get many 32 bit applications, and was thus hampered by the applications released.
It does go one quite a bit in one section about the 16 bit parts of the OS, some of which were replaced in OS/2 2.1, but most of which were there for _compatability_ reasons.
http://www.byte.com/art/9406/sec11/art2.htm
And once again, what does this have to do with the conversation? This is about os/2 2.x, and about a retrofit to that (at the time) older version of the OS to allow for SMP capability. The references to Workplace OS have no bearing on the discussion.... unless you dig.
If you dig about Workplace OS - you find some interesting things, such as the mach-esque microkernel, HAL, and portability to other platforms other than x86. Sounds a lot like the NT kernel to me.
This is really getting old, so let me be Generous and end it.
Just because you can't find it, doesn't mean I can't find a reference to (mostly) shoot down my argument. From: http://www-computerlabor.math.un i-kiel.de/~mressin/os2/history.html "Microsoft decides to shift focus from OS/2 to Windows. IBM and Microsoft split, with IBM taking over development responsibility for all versions up to OS/2 2.0 and Microsoft keeping responsibility of OS/2 3.0, which Microsoft renames to Windows NT (new technology)."
This shows that version os true OS/2 after 2.x were IBM's doing, since they did not have access to the code that lead to NT 3.0. As magmanamous as I am, I cann't conceed completely, for this only shows who had what rights to what technology from the split, not who wrote what parts, or whose source code ended up where. IBM developers were working with microsoft on OS/2 ver 3 when the split happened, and the root of ver 3 was a continuation of OS/2 development, steered by culver to include many enahncements. To say that NT was a complete rewrite and owes nothing to OS/2, and that it is completely unrelated, is just going too far.
Now, the argument is settled, you can continue to believe what you want, as misguided as it may be, and I eat a partial crow. I sugest you spend some time working on your manners - the amount of supositions you stated about me during this debate were rediculous, off base, and rude - I've read your statements in reply to other posters, and really, to be redundant, you need to work on your manners. You seemed to have taken a post that expanded upon a joke someone else made, and attacked it with all vigor. I made the mistake of feeding your troll, and of being too stuborn to let the issue drop - you made the mistake of just being to tightly wound and acting as if the world revolved around your falable memories.
Slashdot Mirror
The media (because they don't understand security), and certain biometric firm executives and salespeople promoto this view - it is a quite horrid view as well. Every engineer and developer I met in thefield (and I met quite a few) thinks it is disgusting - but the media always wants to interview the executive or marketing guy in the suit - not the people who know how the things work and should be used.
Uhm, I am confused, did you read my post and it's description about how multiple layers of security are really needed? - Even Schneier agrees with this principle - it is the use of a biometric instead of a password, which I explicitly mention is not smart, that he is against, and that he thinks adds a false sense of security.
You seem to assume in this post that a biomtric be used on its own, and then discuss failure rates that are low (for authentication), but about right normally for identification (except for facial recognition systems, where it is lower).
You then go and claim that 100 people would be 'accused' of being terrorists, when my text clearly discusses how that is the Wrong way to use such a system - that each claim of such a system needs to be looked at closer, but the system is not to be trusted to be accurate - hence using another camera to zoom in on possible match's face, and get a better reading. The vast majority of a system's failure rate is due to the quality of samples taken.
It is the improper use of such systems, as I described, that leads to being being falsely accused - and from that, we get horrible impressions of the systems themselves, not the (untrained, illinformed) people who run such systems. It is very easy to a security cheif who is getting in trouble because his team is falsely accuwsing people to point to his new facial scannign system and go 'it sucks!' and quit using it, rather than to look into why 'it' isn't working properly and find out that his people don't know how to utilize such a system.
Do you honestly think A bank would use such a system? In my post, I was speaking of reputable systems, not the crappy 200$ versions bought for price over features.
c id=7655 259
As for subscribing, I already do - and having been a developer in the industry for years, I have been reading up on security, as well as holes in biometric systems, for quite some time. But just because device A from a year ago had issues, does not mean new devices will - since any manufacturer who wants to survive will avoid the mistakes of the past.
As for your assertion on facial recognition problems - one system common in use != 'Most facial recognition systems CAN be fooled by holding up a picture'. Usualy, such problems are incurred via the source of the image, when using facial recognition systems that is. Visionics 'FaceIT' for example, becomes less and less secure with cheap cameras. The cost involved in improving such systems is actually very low - the difference between usign a webcam as the image source and using a 2+ megapixel camera. That is 100$ - that does not limit the usage in military bases by any means.
Lighting and other proper setup features are usualy just as important for a useful system.
As for authentication versus identification issues, especially in airports, see my post here:
http://slashdot.org/comments.pl?sid=88410&
When is the last time your employeer knew your back account number, or more accuratly, knew the card number on yoru bacnk card?
your point about a false sense of security, and the need for multiple layers of security in an authentication scheme is correct, but so much of the rest of your post is incorrect, so I feel the need to interject.
- a stolen biometric isn't useful except agaisnt the same sort of scanning system - as in, the same manufacturer. No standard data format exists.
- the pin example is a bad one - the theif needs your card as well (as it is the other layer of security in the system). Anyone who gets the biometric data from the thief will have a hard time using it if they also need the new, shiny, replaced bankcard.
- most biometric systems can tell the difference between dead and living tissue - although this might not stop an ignorant criminal in the first couple of years, it would become commong knowledge that the cut-out-the-eye trick doesn't work once some people ar behind bars.
I think you forget that in any case of security, another weak leak is relying upon any single means of identification/authentication.
The example you gve of the database being compromised is a very horrific one - but the same sort of problem exists for passwords.
I should probably mention that, jsut as passwords are usualy stored as one-way-hashes of the actual plaintext, so biometrics are stored as one-way processings of data taken from a scan, so in either case, the original data (the plantext/scan) isn't being stolen, its representation is being replaced.
Your scenario would be deadly for any authentication scheme, even one that used multiple types of authentication (as all truely secure systems do). Normally, this would be two or more of the following types of items:
- something kept (a security card, id card)
- something known (a password)
- soemthign intrinsic (a biometric, a dna scan)
But all three fail if the root system is compromised that contains the information they are being compared against...
So, your scenario is a problem, but not just for biometrics.
Ok, first, your post seemed directed at me, but was a reply to the parent.
Secondly, how can you claim no ROI for such a rollout? If the return on investment is enhanced security, and if such a system were to provide you with better results than just having extra officers standing around with sheets containing photos of known bad guys, then the ROI is positive.
However, the bad guys must be identified Before they came into an airprort or some other area with such a system - otherwise the system isuseless, as would be name-lists and security guards and wanted posters.
Pre-9/11, most of the hijackers were not on the rader of law enforcement. That sort of failure can ruin investments in all sorts of security technologies - not just biometric ones.
Your idea has problems for several reasons:
- biometric data is not stored as a simple image. It's not stored as a compressed image, or a md5 of the image. It is most often stored as a series one-way-hash values, each of which is derived from some characteristic inherint in the scan. Someone could steal this data, but creating the original image is near impossible, like breaking a 100 kilobyte rsa key.
- biometric data is stored in a different format by every manufacturer. There is no standard - heck, they can barely get a standard API for how to interface with the hardware and drivers (www.bioapi.org), let alone agree on a standard format. Thus, if visa were to start using scanners, and your fingerprint scan were stolen, only visa systems would be affected.
- most authentication systems (other than the implied example of logging onto a computer) use multiple pieces of information, usualy two or more of the following type:
- something remembered ( a password or pin)
- something kept (a security card, a credit card)
- somethign intrinsic (a biometric)
Now, how useful is that fingerprint scan if the visa card it's associated with is not in the theif's hands? How useful is it if you cancel your card and get a new one?
- if someone did manage to steal an image of your fingerprint or retina, it won't do much good: systems these days are able to tell the difference between a dead/living finger, a photo, and even a plastic mold (many systems look for temperature of what is scanned, and can even look for capilary blood flow).
- if someone gets access to a computer system where they can use the information stolen and bypass the scanning device, well, you have much bigger problems: such a breakin would probably compromise things to the point where they can simulate a positive authentication from the driver/hardware, for any user.
- (this one only applies to fingerprints): you have ten fingers, use a different one. For eyes, switch eyes.
Having said all of that, please realize that biometrics are intended to enhance security by adding another layer to the authentication systems in place, not to replace them. A bankcard+pin+fingerprint is more secure than a bankcard+pin.
Anytime you hear/read the mass media promoting the death fo passwords via biometrics, realize that either A) the reporter doesn't get it or B) they have talked to a marketing person at one of the manufacturers who is (most likely in my experience) pandering to the media in an attempt to grow the market and get sales, despite the falsehoods involved.
By the same token, anyone who tells you a password by itself is secure, is also wrong.
AS someone who worked at a company that tried, pre-bapi, to write our own abstraction layer over multiple vendors' drivers, I feel your pain - our entire biz model was to write our own applications in such a way as to be used with different devices, and provide our sdk for other developers to do the same. The more applications, the more demand there would be for the devices - just like video cards with 3d acceleration and games.
What we found was telling in several ways:
- Device manufacturers saw their SDKs as profit center, and did not want to give them away - even though it would promote device sales.
- The engineers at many firms were very, very intelligent about biometrics, but not about software - they would buy driver writing kits for windows and go from there. Sometimes they were very smart electrical engineers as well, designing not only the authentication schemes and putting cameras into little plastic boxes, but designing their own special circuitry to be used on a pci card to provide a high bandwidth interface to thier device, etc. But once again, after going to all that trouble, the firms used driver writing kits.
Of course, what you seem to be missing, is that these vendors have (what appear to them) to be good reason to Not provide open source drivers - or binary drivers for an OS that is run by users who will just try and reverse engineer them.
- often, the actual authentication mechnisms were performed in sofware, with the device just gathering data. Thus, by providing drivers they would open up the core of their product and simply become hardware makers.
- there is no economic incentive - not only does the above point illustrate this, but the demand hasn't been there from the people who actually buy the systems.
Having said all that, note that some closed-box systems that do not require an external computer are rumored to run embedded linux variants...
I think you need to look into security principles. As you say, a lone password is easy to compromise, so is a lone biomtric. However, any truely secure system needs to use multiple forms of identification - preferably two or more of the following:
- something intrinsic (a biometric, dna scan, etc)
- somethign known (a password)
- somethign kept (a security card)
By having more than one step involved, the system is much more secure than any individual part. Somesteals your backcard - but do they have your pin? Or, someone sees your pin - but do they have your card or account number? PINs are actually very simple and easy to break (thoeretically), but are pains to break in reality because of the Other required piece of the puzzle, the bankcard, and how false authentications lead to the removal of the card (most ATMs shred your card after a few false PINs are entered).
similarly: Just because someone steals your face, how will they get ahold of your new bankcard?
After that fact comes the fact that most biometrics are hard to fake - fingerprint scanners these days can be made smart enough to check the temperature of the item placed on them - and some are even smart enough to look for normal temperature differences and gradients within the skin surface, and refuse authentication to 'fingers' that are too regularly or irregularly warm. Some very high end systems look for capilary blood flow... Most facial systems are smart enough to refuse a photo held up of your face, and carrying around a stiff 3d mask of someone's face is kind of obvious.
Also, the fact that every type of scanning device on the market practially has a different data format for the biometric data (which is all one-way, you can get the data from a fingerprint, but not the other way around), and spoofing the data becomes more restrictive - a spoof of, say, visa's system wouldn't work against mastercard's (unless they were using the same equipment).
Having said all that, I'd still like it to be pin+card+face/fingerprint rather than card+biomtric. Biometrics should be used to Enhance security, not replace known or kept-item security methods.
Facial recognition is only 1 of the technologies involved in biometrics... To claim that the whole industry has failed to grow because one Type of biometric does not function well is untrue.
Besides that, your numbers are wrong... facial recognition systems can actually have failure rates higher than that under less than ideal ircumstances, and when put into use as identification, not verification systems.
First, definitions, for those who didn't read the article:
Identification: determin from a scan who someone is, searching over a list of possibilities.
Authentication: determin with reasonable confidence that the user is who they claim they are.
Authentication is much much easier to get right, since you can always ask for a rescan if you are unsure. Authentication systems are designed so that the device (hardware and software) return a confidence level - sometimes a percentage. It is up to the application developer to determin just how high a confidence level you want. If you set it too low, people with similar faces might be abel to authenticate for each other - borthers for example. If set to high, then slight (natural) variations in a person's face can cause rejections. Generaly, you must strike a balance between false positives and rejections. Such a compromise is acceptable, if you have other security measures in place (see note at end of post).
Identification is much, much harder. First of all, it is very cpu intensive - one can model identification as a low-confidence-level authentication against every listed person in the database. If you have 40,000 people in the database, this can take awhile. Hashing doesn't help much, and is illadvised, since we are looking for a close match, not an exact. Biometric data isn't the kind where you can take the first 5 bytes and dump into hash buckets either - but I digress. So, how do you speed it up? You reduce the dataset by reducing the detail in the data you store for each person.
Then you run into the problems with how these systems have been rolled out - using low resolution security cameras is not a good way to get an accurate scan of a person's face - especially when the people being scanned a re small enough (in relation to the scene) to be only 10s of pixels wide.
So, now we know the technical difficulties - but why the bum rap, and why would a police force choose to roll something like this out anyway? This is several fold, but the main thing it comes down to is misconceptions about what these systems are doing, and badly written systems. Due to the limitations mentioned above, these systems can only provide possible matches, like 'Person X is a 20% match against Osama Bin Laden'. the system isn't claiming that the person IS Osama, only that the face appears somewhat similar. As such, the system is supposed to be used as a guide - if it picks someone out, that person deserves more attention - that attention could be a remote-controled security cam singling them oout for a better scan, or for officers in the area to walk over for a better look. Unfortunatly, just because that is how the system is supposed to work does not mean it is used that way - all too often these are rolled out as a way to 'increase security while retaining a minimal police/secuity force'. You get officers who think of a potential match as a authentication, and they send officers running down at high speed only to find it's not Osama... The next potential match they are more hesitent about, and so on, until they mistrust the system completely. Is the system doing anything wrong? No, its that the users don't understand what it is doing. Better training would help, but so would the people making the purchasing descisions understanding the technology, and staffing accordingly.
In the sort of rollouts described above, facial recognition has a success rate of less than 30%, much lowe r than what you describe. With rates that low, people complain, and stories get published. Used properly, the data these sy
The economist article fails to mention the other major reason these systems have not taken off - comparability.
Or, I should say, the Lack of it.
Each fingerprint device on the market uses its own format for storing it's data - making each device incompatible. At first, this would seem to be an easily surmountable problem - but then you must realize that until recently, Every device on the market had its own API for development.
Let me give you an example to illustrate this issue: company X has 2000 employees, and it goes to look at biometric systems - they are either faced with the choice of paying for very expensive equipment from 'long time players' in the industry - who would be around in 2-5 years when the devices start failing due to wear and tear - or choose from some of the 'upstarts', and risk being out in the cold if the company they choose isn't around in several years. a hardware switch down the line not only would incur the cost of re scanning everyone, but the application itself would need to be modified to work with the API for the new device.
Enter the BioAPI (www.bioapi.org) - which proposed a standard api - now widely adopted. You may notice that the Bioapi page mentions it was founded in 1998. It has taken several years for this standard to come to the foreground and there are still roadblocks - not all manufacturers participate freely.
As an example: one rather large manufacturer, Identix (www.identix.com) seems to have been stonewalling for years. Why would a manufacturer do such a thing against what is good for the industry? Because they were leading the industry. When you have all of the high end government contracts coming your way, a standard the opens the doors for the little guy is a Bad Thing for your business - or so they thought.
Take a look at the members list on the bioapi site - identix is listed - then take a look at the supported devices list... not a single identix product.
In 1999 I witnessed this stonewalling firsthand at a meeting in washinton DC. This meeting had manufacturers and interested parties from all over the globe in attendance, including representatives from the US military. The whole agenda for the meeting was how to promote/define standards so that the industry could grow.
I had the unfortunate luck to be seated next to the Identix representative. He had apparently flown in just so he could stonewall - every opportunity he got, he grabbed the microphone and ranted about how we should let the free market dictate standards - that they would come about naturally in the free market (he loved the term free market).
Meanwhile the rest of the group was discussing issues about how to resolve device inter operability - even so far as to discuss how data could be shared between devices. No concrete decisions were made at the meeting, but it did get people talking.
Anyway, my whole point is, one of the major reasons the biometric security industry hasn't grown (as fast as has been predicted for the past 8 years) is because without standards no one wanted to invest in writing applications. It was just too risky.
Note: I am flipping a coin as to wether to post this anonymously or not, since Identix could decide to try and silence this sort of talk...
Wow, that rocks.
Perhaps they got even more flexible since this protability thing was coming up and they knew they needed to be competetive.
Try calling back and speaking to someone else - I know I changed two of my plans within 4 months of starting two year agreements.
I know that the new verizon specials are much better than the old plans - my 75$/month plan is now a 60$/month special.
The special is so good, I'm moving my other phones to it - previously that had 50$/month plans plus a 10$ feature-add (mobile-to-mobile minutes). The 60$ plan includes the mobile-to-mobile minutes and 400 more peak minutes than the old 50+10 I was doing.
So, I'll save 15$/month, and get 400 extra peak minutes on two of my phones. I'd hoped for even bigger savings, but 180$/year with more minutes isn't too shabby.
Now, if only I got good reception in all of my house - I'm apparently located at the edge of two cells...
Are you using Verizon?
I know verizon lets you switch plans in mid-stream, as long as you get a new term on the selected plan that is longer than what remains on your current agreement. IE, if you have 14 months left, you have to get a new 2 year agreement.
I've never had a problem switching to new promotional plans since I got my phones (I have 3 phones in my name for myself, my wife, and the business - no land lines). This may just be a Verizon thing though.
Ah, but that is the problem itself - the poster I was replying to did not state 'there are problems with Darwins theory of natural selection' - he stated that there are challenges to evolution - which, although described by Darwin's theories, is not exclusively Darwin's. I know, I know, it comes down to semantics and what did the poster actually mean, but thats why I commented in the first place: because his phrasing would imply to the casual reader that evolution itself was in question - when in fact, it is only theories about how evolution works that are being questioned. Am I making sense? Maybe I was seeing a mountain where there was a molehill...
What you describe exists, but alas, it runs windows...
t m
And is expensive for basicly a suped-up pda with a terminal services clients.
Look here: http://www.viewsonic.com/products/smartdisplays.h
If they were not quite as expensive, I'd buy one in a heartbeat.
Note: They use 802.11b to communicate with the host machine, and I do not know if they will work remotely very well.
If you want to save money, look on ebay for the predecessor to the, the viewsonic 'superPDA' - its basicaly the model 100 (see the business version) airpanel, but with a pda version of CE used instead of a stripepd down version for 'smart displays'.
Oh, and as for a slideshow while they recharge... uhm, I guess a screensaver?
Even now, there's substantial logical and statistical problems with the "proofs" of Evolution.
Are you refering to darwinian ideals of evolution, or the concept as a whole?
True original darwinism as the sole motivator for the changes in species over time is being challenged, but the concept as a whole - that life came from very simple beginings and has changed/adapted over time is not. The mechanisms involved are what are being challenged - such as the idea that small changes in genotype over time that favor the survival of a particular subset of a species lead to massive changes in the long-view. Fossils for the 'in-between' variants are not being found, hence it is becoming more widly accepted that large leaps are made, and that such large leaps could actualy be triggered by environmental pressure.
However, these new mechanisms being discused and discovered are just that - mechnisms. Evolution as Darwin envisioned it may be being disproven, but the idea that life evolves over time is not.
If, instead of refering to darwinian evolution, you are refering to evolution as a whole - then you are seriously mistaken. There is no creationist or other theory of life that is being pushed ahead of evolution by scientists. The logical and statistical problems you mention are about the problems with darwinian evolution and its mechanisms.
The bug you describe with 1.4.x and swing components is a known problem with the 1.4.1 series - though it seems to be fixed for me in the latest 1.4.2_02 release, and has been fixed since the 1.4.2 initial release.
The workaround is to disable direct3d and directdraw when using swing - they apparently do some flaky calls to directX libraries in order to speed up the gui drawing, and the ati drivers _hate_ it.
I'd refer you to the bug report on sun's site, but it requires registration, so I'll just cut-n-paste the workarounds as found on a non-sun site:
"
Upgrade to Java 1.4.2
-or-
Set the Java command-line flag -Dsun.java2d.noddraw=true to disable these direct draw optimizations in Java 1.4.1.
-or-
Switch back to JVM version 1.4.0 which does not have the DirectDraw optimizations which trigger this problem in the ATI driver. This is a pretty undesirable solution, as many bugs were fixed in 1.4.1 and simple things like TreeTable fail to work in 1.4.0. Be sure to use Java Web Start v1.0.1. Web Start v1.2 is only installed with JRE 1.4.1.
-or-
Goto "Display properties -> Settings -> Advanced -> Troubleshoot" and drag the Acceleration slider all the way to the left to turn off hardware acceleration. (this will reduce but not eliminate the lockups)
"
How is the parent, an obvious Joke if you follow the link, flamebait? if you don't like the humor, thats fine, but flamebait is a post designed to instigate an argument - I don't think anyone with 2 brain cells would say a silly joke is flamebait...
Wait until the RIAA gets a hold of this news, they'll sue the pants off that puny entity known as the ITU, and These guys wont even get a penny.
Ah, I took your mention of SAG to mean you wondered if they knew SAG actors were having their works reused, not the way you intended, my bad.
Quote:
Thought Equity gets agreement from any and all parties to get exclusive worldwide distribution rights, Schaff said. Thought Equity also doesn't use ads featuring actors who belong to the Screen Actors Guild.
I am the troll? I am trolling?
n i-kiel.de/~mressin /os2/history.html
I made a joke, expanding upon someone elses joke. You didn't take it as a joke, and you went off, YELLING in your response... but I am a troll?
I don't think you know the term properly.
Really, I mean, these links are all mostly off topic, and none of them address the issue directly of wether OS/2 ver 3 became NT. Your statements to the contrary in your post, and your (once again) yelling, do not disprove the basic fact that the two _are_ related and came from the same original project.
I mean, come on, does this link:
http://tunes.org/~unios/oskernels.html
Has anything to do with the conversation other than to prove you would rather read a single paragraph description of what a microkernel is rather than read the actual source data? The books I provided links to on amazon are all college (and in some cases, graduate course) textbooks - I'm pretty sure they define a microkernel a little bit more in depth then what this link does.
Or did you think this link would 'teach me' about NT? This link provides no comparison between the NT kernel and the OS/2 kernel, and thus is useless in this discussion other than to be redundant.
http://pages.prodigy.net/michaln/history/os220/
This, once again, has nothing to do with the conversation - which is about OS/2 ver 3 that was underdevelopment when the MS-IBM split happened.
Let me sumerize this page for you:
OS/2 2.0 was 32 bit, but didn't get many 32 bit applications, and was thus hampered by the applications released.
It does go one quite a bit in one section about the 16 bit parts of the OS, some of which were replaced in OS/2 2.1, but most of which were there for _compatability_ reasons.
http://www.byte.com/art/9406/sec11/art2.htm
And once again, what does this have to do with the conversation? This is about os/2 2.x, and about a retrofit to that (at the time) older version of the OS to allow for SMP capability.
The references to Workplace OS have no bearing on the discussion.... unless you dig.
If you dig about Workplace OS - you find some interesting things, such as the mach-esque microkernel, HAL, and portability to other platforms other than x86. Sounds a lot like the NT kernel to me.
This is really getting old, so let me be Generous and end it.
Just because you can't find it, doesn't mean I can't find a reference to (mostly) shoot down my argument.
From:
http://www-computerlabor.math.u
"Microsoft decides to shift focus from OS/2 to Windows. IBM and Microsoft split, with IBM taking over development responsibility for all versions up to OS/2 2.0 and Microsoft keeping responsibility of OS/2 3.0, which Microsoft renames to Windows NT (new technology)."
This shows that version os true OS/2 after 2.x were IBM's doing, since they did not have access to the code that lead to NT 3.0.
As magmanamous as I am, I cann't conceed completely, for this only shows who had what rights to what technology from the split, not who wrote what parts, or whose source code ended up where. IBM developers were working with microsoft on OS/2 ver 3 when the split happened, and the root of ver 3 was a continuation of OS/2 development, steered by culver to include many enahncements. To say that NT was a complete rewrite and owes nothing to OS/2, and that it is completely unrelated, is just going too far.
Now, the argument is settled, you can continue to believe what you want, as misguided as it may be, and I eat a partial crow. I sugest you spend some time working on your manners - the amount of supositions you stated about me during this debate were rediculous, off base, and rude - I've read your statements in reply to other posters, and really, to be redundant, you need to work on your manners. You seemed to have taken a post that expanded upon a joke someone else made, and attacked it with all vigor. I made the mistake of feeding your troll, and of being too stuborn to let the issue drop - you made the mistake of just being to tightly wound and acting as if the world revolved around your falable memories.