Note to most/.er's go ahead and ignore the below...
Dr. Pande-
Great study (see point number four of Good CS, Good Chemistry for what I particularily liked about it). I'm about to ask a bunch of scientific questions in the worst possible forum but what the hey...
I really like intelligent application of physical chemistry here (specifically why you can do this with only a 20ns simulation). It seem obvious now (after reading the paper), but for some reason I'd never considered it, very clever. Three questions.
1) Are you surprised that it worked at all? What with implicit solvent and the like I was surprised with the degree of success that you had. Specifically, it seems the usually important water entropy term would be next to impossible to account for. Not to read too far into the study, but do you thin this may mean that perhaps water entropy helps with stability (pretty hard to argue against these days) but has little effect on the forward rate in this case (ie. it affect mostly the unfolding here)? Or did I miss a subtle hydrophobic exposure penalty or something? I'm a bit out of date here, is 'generalized Born/surface area implicit solvent' how this is taken care of?
2) You mentioned simulation of Abeta in another comment here. Did you catch the entertaining coincidences at last year's Biophysical Society? Specifically, two adjacent posters on Abeta. Someone from Ruth Nussinov's (I think it was DZ Gomara, but my memory may be lying to me) group got a structure of Abeta in the fiber by simulation which was surprisingly close to the structure that AT Petkova of Tycko's had by SSNMR. Both were rather preliminary but it was fun to see.
3) Are you going to try and extract information from the trajectories themselves. It seems like you're just trying to cast everything in terms of an observable at this point (ie. paying particluar attention to only those trajectories that produce a structure close to your expected conformation). It seems that there should be some underlying information regarding the actual folding landscape (if you believe that stuff) in the trajectories them selves. I'm not sure how you would do it, ie. I can imagine how to make rules regarding local structure and then see what happens to the trajectories, but the reverse seems hard. Just wondering if you're after that sort of information, given that it seems like you might have a data set which actually permits it:)
Okay, okay, back to work...
S
Given that you've almost stated the Levinthal paradox I'll assume you're familiar with it, but missed the point. Basically, it states that even in the simplest description of protein conformation (say 3 possible states each for 100 amino acids) can't be searched in a reasonable period of time, the shortest feasible time that a protein could sample a state in about 10^-13s. This works out to be ~10^27 years to check all the 3^100th states (borrowing Styer's description of this). This is clearly wrong, proteins fold in milliseconds (okay ns-100s of but you get my point). The clear conclusion is that proteins don't sample every conformation availible, or even any singificant fraction of them. There must be some fashion by which frequent short range and random long range contacts guide the protein into a 'pathway' of folding.
The nifty thing with the folding@home study is that there were able to basically show that invoking a simple physical force field system was enough to get pathways, though they don't make too big a deal about this, maybe someone else has already done this, but I'd be surprised if they managed to do as many trajectories as were done here. I imagine it'll be a while before they process the trajectories to try and find actual pathways (very compute intensive), but the fact that they found a comparable rate (we're not talking global conformation here, these are kinetics) suggests that they may be sitting on top of an actual description of the folding pathway for this teeny protein. Spiffy!
Well, it's not a team as such right now, but I've got an idea that could make use of that sort of idle time. Biologically relevant (though not all atom), tedious for a single run and only interesting after about a thousand or so cycles. Ie. readily parallelized. Could probably have it ready to test out in early spring. Does that mean you've got 30k CPUs running with nothing on them or are they all powered down? If you're interested reply to this and then we can hopefully come up with a way of getting in touch without having to broadcast our email addresses to the world.
Alternatively, who ever is following up on folding@home (I guarantee this is just the beginning) would be thrilled with access to that kind of power.
That's pretty good. Lab to lab variation when repeating measurements like this is usually this bad or worse. Factor of two would have been reasonable. When comparing in silico work to experimental its usually considered good when you're within a factor of ten. More commonly a series of related proteins is ranked in order (by some property) and then compared to experimental measurements of said rankings.
Also, keep in mind that this is a microsecond folding rate. Rates in the ms regime of folding are routinely measured with high accuracy, the microsecond regime is really hard as it usually takes longer than that to do what ever you're going to do to trigger folding in the first place. The number in the paper is 7.5us +/- 3.5us, so they got as close as could be expected.
First, you can't expect to go from no success to complete success overnight. People have been trying to fold proteins for some time now and have basically failed because it is freakin' hard. The theory is in principle in place, a least to a first approximation, but the calculations are so intensive that they have basically beaten every comer. As an undergraduate I remember how everyone in the field thought getting bigger and better grants and buying bigger and bigger computers was the answer. Oh to be SGI in those days. They sum up the problem pretty well in the Nature paper, essentially a modern (desktop) computer would require a few decades to crunch through a single useful length simulation. Then you need to do it many times to get a useful answer (say 100-1000). Even supercomputers are going to balk at that kind of calculation. Moore's law what it is, we should then be able to get through an in silico simulation in a week on a single computer (when its this fast crystallography really will be dead) by, oh say 2040 at best. (somebody want to calculate that exactly, 10000yrs -> 0.02yrs is how many doubles). So yes, this hasn't gotten rid of x-ray crystallography just yet.
But this is still really cool. Complaints about interface and maintenance aside, this was a great system. It relied on four pretty bright insights.
First, that distributed computing is essentially the poor man's (cough, the academic's) super computer. Also, it automatically adapts itself to technological improvements. People will buy new computers from time to time and, hopefully, reload your software.
Second, that there was no reason other than no one had sufficiently brute forced the process that the existing methods shouldn't work. They use a bunch of 'cheating' techniques to make this managable during the screen saver timescale, such as a united atom model (I think that means they ignore aliphatic hydrogens) and implicit solvent (don't treat it as individual solvent molecules, just a uniform field). It was an open question as to whether this approach would work at all or if you had to go over to much more explicit methods to get it to work at all. It appears that this has kinda worked with the cheater methods in place.
Third, choice of a test case. Yes they chose something that was small. This isn't surprising. They wanted to be done sometime this decade, remember there is a graduate student as the primary author here. Small was necessary. However they also chose a FAST-FOLDING protein. That was clever. Basically, even with distributed computing, it is still hard to simulate a full microsecond of time. Thus they chose something that had some chance of completing its folding one the time scale that they could look at.
Fourth, they remembered their P-Chem. It is really hard to run these things to completion... so they didn't. You don't have to run the simulation until 99% of the molecules have completely folded, just until an appreciable number have folded and you can extrapolate the behavior from that. They ran a 20ns simulation (at the longest). The thing takes 7us for ~60% to fold. As a result only once in a great ong while did one of the simulations actually produce a folded protein. But by doing it ~10000 times they could figure out how that translated into the rate constant. That's clever.
That said, yes there is a long way to go on this, but its still a really clever paper. No we haven't cured cure cancer yet, but its still progress. And forget an in silico structure of the ATPase, that's largely understood already (check the RSCB/PDB there's a bunch). The real challenge will be getting a structure that size that hasn't been solved by other methods and convincing anyone else that you're right!
Disclosure- I don't have PhD in this area yet, but I'm close.
Actually they did review one Antec powersupply (the True380 I think). It came in just behind the leaders, though not for any reason I could tell from their spec's. Seems to have performed at manufavctures specification and was reasonably quite. Maybe they didn't think it was quite as good a deal. It was nice to see one of the readily availible brands come in near the top though.
Actually, I was curious, if you're using a lot of Antec supplies could you tell me what the practical difference, if any, is between Antec's True power supply line and their Sl line? Is it just that the tolerances on the voltages are a little tighter?
If you want fast memory (which is often useful) but don't feel like shelling out $$$ for the greater than 2x cost of RDRAM, DDR333 is now (actually it was long before this article) an option, even supported by Intel. In some cases it even outperforms the rather expensive 1066 RDRAM. It's been around for a bit from the other manufacturers but this seems like an attempt by Intel to stick a fork in RDRAM despite the fact they were the ones championing it all along. Cheaper & higher performance = good.
The rest of it is just a bunch of bells and whistles. When will it end, probably never.
Even worse, look at the far left of the picture of the ports on the board, there is a Firewire port, near as I can tell. Top of the bank to the right of the PS2 connections. Looks an awful lot like 2 USB and an IEE1394...
nVidia has that covered with its planned boards, its called an AGP slot:)
From a press release on nVidia's site regarding nForce2 chipsets In addition, the inclusion of an AGP 8X expansion bus guarantees a constant upgrade path for the industry?s fastest external add-in cards, for even more top-of-the-line graphics processing power.
True, you might not find 5.1 sound or AGP8x/GeForce4 4400 performance,
Okay I have no clue what I'm talking about here but my guess is that an integrated GPU would have pretty fast communications with the rest of the motherboard, afterall I think they share the system RAM, this could well be faster than AGP8x, which currently does next to nothing according to Tom's Hardware Guide is basically a marketing feature anyhow.
Hate replying to my own post but I should point out that I couldn't read the article (they know they've been posted here). I can't tell if the geFORCE 4 mobo is mentioned or not.
Sure, it still may not be enough for you 'hardcore gamers'...
but Shuttle computers and nVIDIA are planning on releasing an integrated GeFORCE 4 MX motherboard. This will be particularily cool for shuttle itself, who makes relatively small (and attractive) barebones systems. Not having to leave space for an AGP card will help them a lot. (btw- I have nothing to do with either of these companies)
Also, I don't think the purpose of these integrated cards is generally to keep gamers happy, they'll want to upgrade every few months anyhow. Integration is there to make it cheaper for the rest of us to get decent graphics on a cheap box.
School are perpetually complaining about a lack of funds for computers. In a way the PS1 is a really cheap computer that the schools can afford to buy in quantity and loan to kids. What does a PS1 go for these day $50? Need a monitor but most families have a TV. If a school buys enough of them I'm sure the price would go down. Media... CD based. School buys a general license on a piece of software/buys a package to author their own and prints a hundred copies using an off the shelf CD burner. Add one of those chips to allow any CD to be used and you have a really cheap computer solution for students. Granted the input (gamepad) is a bit limiting, but I imagine a keyboard already exists or could be built if there was sufficient demand. Alternatively, could one just come up with some sort of converter for a PS2 input? Probably not.
Pretty clever really. Thinking back, I wrote all my high school essays on an Atari with 512k and a decent word processor, I think the whole thing may have booted/run software off a 3.5" floppy and an internal ROM. Didn't have a spell checker or anything but considering I was writing for highschool I think spell checkers may have been a bad idea at the time.
As far as the 'edutainment' gripe is concerned, how about having a seemingly meaningless code output at the end - similar to the old NES games continuation codes. Student writes it down and gives it to teacher, teacher uses conversion program to find out that the student has missed 65% of the questions, mostly the hard ones.
What would be really, really cool is if there could be a way of hacking together a cheap solution for the output problem. Student has CD with Linux port - exists for PS1 if I recall - some mini gui and a small office solution. Student writes essay. Are they SOL if they want to say print/email the text to teacher? Is there an obvious output solution I'm unaware of? I can think of that wouldn't at least triple the cost of the system.
You don't need to know all the possible answers, a few thousand should do. And if you choose the queries randomly, it would be about as good as knowing all the answers to billions of questions, as far as a hack is concerned. More efficient to just measure a few query/speckle patterns directly.
As far as modeling the suspension of sphere in a computer and using that to pretend to be a vendor asking for a verification, sure, but you would need to actually get someone's fob thingy and map, very carefully, all the positions of the sphere, which is not trivial. Specifically, it would require a significant ammount of time. From a crime standpoint, you would be better off just using the fob you've obtained (stolen) directly on a normal scanner.
And of course, as has been pointed out many times above, this is assuming that the 'secure server' really is.
Actually, using it as an optical signature should work now with the fob technology as described.
The company who wants your signature just requests a validation on your resin/fob thiny. The difference is that instead of tossing the query/response as usual, they keep it. The queries are used only once by the validation company so that wouldn't be a problem. Futher, the validation company would probably keep a record of who asked for validations and what query/response went with it, so they could be an independent party to verify that yes you did let so and so check your resin thingy at such and such a time on such day.
The only really big problems here are the (a) scanners will probably be expensive and (b) if it gets stolen you're in trouble if you don't tell the verification company ASAP. But those were both problems (and b continues to be) when credit card were initially introduced.
does this mean the issuer is going to have to store a terabit of data for each user?
The issuer doesn't have to worry about knowing ALL possible query response combinations, just a number of combinations that is large enough to last the fob's expected lifetime, maybe 1000 queries (angle, position etc) and responses (the hashed specle outputs). This is still a significant amount of info, but should be less than 10M for each fob, a reasonable amount by today's standards.
All an attacker has to do is figure out what the subset that will be used is.
A good point, but since the issuer controls the questions, they could randomly decide which queries out of a very large set to store. A theif would then need to know the answers to many, many more queries than the issuer needs to store.
Insightful! Yes...
I went in and read the actual article (in Science, with subscription, sorry), as a result, here's a rather verbose response.
You're pretty close to what the original authors actually propose in the article. Essentially, the fob is just a rugged, cheap, light weight way of carrying around a zillion answers of ridiculous complexity to a whole bunch of simple questions.
Before you're given a fob they would scan it at every angle, position and wavelength of interest, generating an enourmous number of possible questions to ask. Then they store the answer to all the questions. When you actually use the thing to make a purchase, a question is asked (ie. what do i get if I illuminate at X angle, Y position on the resin and with Z wavelength). A particular answer is given and compared to the stored answer. If it agrees, great. If not, try another. If it fails again, then it doesn't validate. The key thing though is that questions are never asked twice! As a result, the questions and the answers could be intercepted and stolen one by one and it wouldn't matter, as they could never be used again! When they run out of questions to ask (or get close) they have you get another 1 cent fob.
The only real security problem I could imagine would be if someone cracked a reader and had it try to read all possible combinations while you were standing there. This would probably take too long to make it worth it. A partial read, well the theif doesn't control which question gets asked and if you have too many bad verifications, ie you're trying to use a partial read, they might drop by to check out your reader... Two other problems, if it gets stolen, you're SOL. Second, the reader is likely to be expensive, making it hard to use this to allow purchase authorization at home.
So your problem...
The problem with this is that the validation server would have to know what the right answers are to all of the possible questions, and that creates a problem: either there would be waay too much data stored for each card, or there would only be a limited number of "questions" the server could ask.
The answer, a limited number of questions. This would probably be fine tuned to balance out the replacement cost, anticipated number of validations during the lifetime of the thing etc. Seems like storage might be an issue though.
As far as the half baked work around everyone else seems to be proposing, reading the article helps. The only one which actually might work, reproducing the resin using the paired laser/heat harden resin approach might actually work at some point. But it would require having the fob in the theif's possession for so long that the original would probably have noticed as missing, canceled the old one and gotten a new one by the time it was ready.
Slashdot Mirror
Note to most /.er's go ahead and ignore the below...
:)
Okay, okay, back to work...
S
Dr. Pande-
Great study (see point number four of Good CS, Good Chemistry for what I particularily liked about it). I'm about to ask a bunch of scientific questions in the worst possible forum but what the hey...
I really like intelligent application of physical chemistry here (specifically why you can do this with only a 20ns simulation). It seem obvious now (after reading the paper), but for some reason I'd never considered it, very clever. Three questions.
1) Are you surprised that it worked at all? What with implicit solvent and the like I was surprised with the degree of success that you had. Specifically, it seems the usually important water entropy term would be next to impossible to account for. Not to read too far into the study, but do you thin this may mean that perhaps water entropy helps with stability (pretty hard to argue against these days) but has little effect on the forward rate in this case (ie. it affect mostly the unfolding here)? Or did I miss a subtle hydrophobic exposure penalty or something? I'm a bit out of date here, is 'generalized Born/surface area implicit solvent' how this is taken care of?
2) You mentioned simulation of Abeta in another comment here. Did you catch the entertaining coincidences at last year's Biophysical Society? Specifically, two adjacent posters on Abeta. Someone from Ruth Nussinov's (I think it was DZ Gomara, but my memory may be lying to me) group got a structure of Abeta in the fiber by simulation which was surprisingly close to the structure that AT Petkova of Tycko's had by SSNMR. Both were rather preliminary but it was fun to see.
3) Are you going to try and extract information from the trajectories themselves. It seems like you're just trying to cast everything in terms of an observable at this point (ie. paying particluar attention to only those trajectories that produce a structure close to your expected conformation). It seems that there should be some underlying information regarding the actual folding landscape (if you believe that stuff) in the trajectories them selves. I'm not sure how you would do it, ie. I can imagine how to make rules regarding local structure and then see what happens to the trajectories, but the reverse seems hard. Just wondering if you're after that sort of information, given that it seems like you might have a data set which actually permits it
Given that you've almost stated the Levinthal paradox I'll assume you're familiar with it, but missed the point. Basically, it states that even in the simplest description of protein conformation (say 3 possible states each for 100 amino acids) can't be searched in a reasonable period of time, the shortest feasible time that a protein could sample a state in about 10^-13s. This works out to be ~10^27 years to check all the 3^100th states (borrowing Styer's description of this). This is clearly wrong, proteins fold in milliseconds (okay ns-100s of but you get my point). The clear conclusion is that proteins don't sample every conformation availible, or even any singificant fraction of them. There must be some fashion by which frequent short range and random long range contacts guide the protein into a 'pathway' of folding.
The nifty thing with the folding@home study is that there were able to basically show that invoking a simple physical force field system was enough to get pathways, though they don't make too big a deal about this, maybe someone else has already done this, but I'd be surprised if they managed to do as many trajectories as were done here. I imagine it'll be a while before they process the trajectories to try and find actual pathways (very compute intensive), but the fact that they found a comparable rate (we're not talking global conformation here, these are kinetics) suggests that they may be sitting on top of an actual description of the folding pathway for this teeny protein. Spiffy!
Well, it's not a team as such right now, but I've got an idea that could make use of that sort of idle time. Biologically relevant (though not all atom), tedious for a single run and only interesting after about a thousand or so cycles. Ie. readily parallelized. Could probably have it ready to test out in early spring. Does that mean you've got 30k CPUs running with nothing on them or are they all powered down? If you're interested reply to this and then we can hopefully come up with a way of getting in touch without having to broadcast our email addresses to the world.
Alternatively, who ever is following up on folding@home (I guarantee this is just the beginning) would be thrilled with access to that kind of power.
That's pretty good. Lab to lab variation when repeating measurements like this is usually this bad or worse. Factor of two would have been reasonable. When comparing in silico work to experimental its usually considered good when you're within a factor of ten. More commonly a series of related proteins is ranked in order (by some property) and then compared to experimental measurements of said rankings.
Also, keep in mind that this is a microsecond folding rate. Rates in the ms regime of folding are routinely measured with high accuracy, the microsecond regime is really hard as it usually takes longer than that to do what ever you're going to do to trigger folding in the first place. The number in the paper is 7.5us +/- 3.5us, so they got as close as could be expected.
Well, I kinda agree and I kinda disagree.
First, you can't expect to go from no success to complete success overnight. People have been trying to fold proteins for some time now and have basically failed because it is freakin' hard. The theory is in principle in place, a least to a first approximation, but the calculations are so intensive that they have basically beaten every comer. As an undergraduate I remember how everyone in the field thought getting bigger and better grants and buying bigger and bigger computers was the answer. Oh to be SGI in those days. They sum up the problem pretty well in the Nature paper, essentially a modern (desktop) computer would require a few decades to crunch through a single useful length simulation. Then you need to do it many times to get a useful answer (say 100-1000). Even supercomputers are going to balk at that kind of calculation. Moore's law what it is, we should then be able to get through an in silico simulation in a week on a single computer (when its this fast crystallography really will be dead) by, oh say 2040 at best. (somebody want to calculate that exactly, 10000yrs -> 0.02yrs is how many doubles). So yes, this hasn't gotten rid of x-ray crystallography just yet.
But this is still really cool. Complaints about interface and maintenance aside, this was a great system. It relied on four pretty bright insights.
First, that distributed computing is essentially the poor man's (cough, the academic's) super computer. Also, it automatically adapts itself to technological improvements. People will buy new computers from time to time and, hopefully, reload your software.
Second, that there was no reason other than no one had sufficiently brute forced the process that the existing methods shouldn't work. They use a bunch of 'cheating' techniques to make this managable during the screen saver timescale, such as a united atom model (I think that means they ignore aliphatic hydrogens) and implicit solvent (don't treat it as individual solvent molecules, just a uniform field). It was an open question as to whether this approach would work at all or if you had to go over to much more explicit methods to get it to work at all. It appears that this has kinda worked with the cheater methods in place.
Third, choice of a test case. Yes they chose something that was small. This isn't surprising. They wanted to be done sometime this decade, remember there is a graduate student as the primary author here. Small was necessary. However they also chose a FAST-FOLDING protein. That was clever. Basically, even with distributed computing, it is still hard to simulate a full microsecond of time. Thus they chose something that had some chance of completing its folding one the time scale that they could look at.
Fourth, they remembered their P-Chem. It is really hard to run these things to completion... so they didn't. You don't have to run the simulation until 99% of the molecules have completely folded, just until an appreciable number have folded and you can extrapolate the behavior from that. They ran a 20ns simulation (at the longest). The thing takes 7us for ~60% to fold. As a result only once in a great ong while did one of the simulations actually produce a folded protein. But by doing it ~10000 times they could figure out how that translated into the rate constant. That's clever.
That said, yes there is a long way to go on this, but its still a really clever paper. No we haven't cured cure cancer yet, but its still progress. And forget an in silico structure of the ATPase, that's largely understood already (check the RSCB/PDB there's a bunch). The real challenge will be getting a structure that size that hasn't been solved by other methods and convincing anyone else that you're right! Disclosure- I don't have PhD in this area yet, but I'm close.
Actually they did review one Antec powersupply (the True380 I think). It came in just behind the leaders, though not for any reason I could tell from their spec's. Seems to have performed at manufavctures specification and was reasonably quite. Maybe they didn't think it was quite as good a deal. It was nice to see one of the readily availible brands come in near the top though.
Actually, I was curious, if you're using a lot of Antec supplies could you tell me what the practical difference, if any, is between Antec's True power supply line and their Sl line? Is it just that the tolerances on the voltages are a little tighter?
Simple...
If you want fast memory (which is often useful) but don't feel like shelling out $$$ for the greater than 2x cost of RDRAM, DDR333 is now (actually it was long before this article) an option, even supported by Intel. In some cases it even outperforms the rather expensive 1066 RDRAM. It's been around for a bit from the other manufacturers but this seems like an attempt by Intel to stick a fork in RDRAM despite the fact they were the ones championing it all along. Cheaper & higher performance = good.
The rest of it is just a bunch of bells and whistles. When will it end, probably never.
Even worse, look at the far left of the picture of the ports on the board, there is a Firewire port, near as I can tell. Top of the bank to the right of the PS2 connections. Looks an awful lot like 2 USB and an IEE1394...
Hopefully, this guy knows what one is...
nVidia has that covered with its planned boards, its called an AGP slot :)
From a press release on nVidia's site regarding nForce2 chipsets
In addition, the inclusion of an AGP 8X expansion bus guarantees a constant upgrade path for the industry?s fastest external add-in cards, for even more top-of-the-line graphics processing power.
Thought the article was familiar, figured slashdot was just late in reporting it, turns out they're redundant instead.
True, you might not find 5.1 sound or AGP8x/GeForce4 4400 performance,
Okay I have no clue what I'm talking about here but my guess is that an integrated GPU would have pretty fast communications with the rest of the motherboard, afterall I think they share the system RAM, this could well be faster than AGP8x, which currently does next to nothing according to Tom's Hardware Guide is basically a marketing feature anyhow.
Hate replying to my own post but I should point out that I couldn't read the article (they know they've been posted here). I can't tell if the geFORCE 4 mobo is mentioned or not.
Sure, it still may not be enough for you 'hardcore gamers'...
but Shuttle computers and nVIDIA are planning on releasing an integrated GeFORCE 4 MX motherboard. This will be particularily cool for shuttle itself, who makes relatively small (and attractive) barebones systems. Not having to leave space for an AGP card will help them a lot. (btw- I have nothing to do with either of these companies)
Their joint press release
Also, I don't think the purpose of these integrated cards is generally to keep gamers happy, they'll want to upgrade every few months anyhow. Integration is there to make it cheaper for the rest of us to get decent graphics on a cheap box.
School are perpetually complaining about a lack of funds for computers. In a way the PS1 is a really cheap computer that the schools can afford to buy in quantity and loan to kids. What does a PS1 go for these day $50? Need a monitor but most families have a TV. If a school buys enough of them I'm sure the price would go down. Media... CD based. School buys a general license on a piece of software/buys a package to author their own and prints a hundred copies using an off the shelf CD burner. Add one of those chips to allow any CD to be used and you have a really cheap computer solution for students. Granted the input (gamepad) is a bit limiting, but I imagine a keyboard already exists or could be built if there was sufficient demand. Alternatively, could one just come up with some sort of converter for a PS2 input? Probably not.
Pretty clever really. Thinking back, I wrote all my high school essays on an Atari with 512k and a decent word processor, I think the whole thing may have booted/run software off a 3.5" floppy and an internal ROM. Didn't have a spell checker or anything but considering I was writing for highschool I think spell checkers may have been a bad idea at the time.
As far as the 'edutainment' gripe is concerned, how about having a seemingly meaningless code output at the end - similar to the old NES games continuation codes. Student writes it down and gives it to teacher, teacher uses conversion program to find out that the student has missed 65% of the questions, mostly the hard ones.
What would be really, really cool is if there could be a way of hacking together a cheap solution for the output problem. Student has CD with Linux port - exists for PS1 if I recall - some mini gui and a small office solution. Student writes essay. Are they SOL if they want to say print/email the text to teacher? Is there an obvious output solution I'm unaware of? I can think of that wouldn't at least triple the cost of the system.
Of course the point of BattleShip is that you can't see the other players boats, the whole guessing positions thing and all.
You don't need to know all the possible answers, a few thousand should do. And if you choose the queries randomly, it would be about as good as knowing all the answers to billions of questions, as far as a hack is concerned. More efficient to just measure a few query/speckle patterns directly.
As far as modeling the suspension of sphere in a computer and using that to pretend to be a vendor asking for a verification, sure, but you would need to actually get someone's fob thingy and map, very carefully, all the positions of the sphere, which is not trivial. Specifically, it would require a significant ammount of time. From a crime standpoint, you would be better off just using the fob you've obtained (stolen) directly on a normal scanner.
And of course, as has been pointed out many times above, this is assuming that the 'secure server' really is.
Actually, using it as an optical signature should work now with the fob technology as described.
The company who wants your signature just requests a validation on your resin/fob thiny. The difference is that instead of tossing the query/response as usual, they keep it. The queries are used only once by the validation company so that wouldn't be a problem. Futher, the validation company would probably keep a record of who asked for validations and what query/response went with it, so they could be an independent party to verify that yes you did let so and so check your resin thingy at such and such a time on such day.
The only really big problems here are the (a) scanners will probably be expensive and (b) if it gets stolen you're in trouble if you don't tell the verification company ASAP. But those were both problems (and b continues to be) when credit card were initially introduced.
does this mean the issuer is going to have to store a terabit of data for each user?
The issuer doesn't have to worry about knowing ALL possible query response combinations, just a number of combinations that is large enough to last the fob's expected lifetime, maybe 1000 queries (angle, position etc) and responses (the hashed specle outputs). This is still a significant amount of info, but should be less than 10M for each fob, a reasonable amount by today's standards.
All an attacker has to do is figure out what the subset that will be used is.
A good point, but since the issuer controls the questions, they could randomly decide which queries out of a very large set to store. A theif would then need to know the answers to many, many more queries than the issuer needs to store.
Ick-
should have used preview
clearly I'm a newbie
Insightful! Yes... I went in and read the actual article (in Science, with subscription, sorry), as a result, here's a rather verbose response. You're pretty close to what the original authors actually propose in the article. Essentially, the fob is just a rugged, cheap, light weight way of carrying around a zillion answers of ridiculous complexity to a whole bunch of simple questions. Before you're given a fob they would scan it at every angle, position and wavelength of interest, generating an enourmous number of possible questions to ask. Then they store the answer to all the questions. When you actually use the thing to make a purchase, a question is asked (ie. what do i get if I illuminate at X angle, Y position on the resin and with Z wavelength). A particular answer is given and compared to the stored answer. If it agrees, great. If not, try another. If it fails again, then it doesn't validate. The key thing though is that questions are never asked twice! As a result, the questions and the answers could be intercepted and stolen one by one and it wouldn't matter, as they could never be used again! When they run out of questions to ask (or get close) they have you get another 1 cent fob. The only real security problem I could imagine would be if someone cracked a reader and had it try to read all possible combinations while you were standing there. This would probably take too long to make it worth it. A partial read, well the theif doesn't control which question gets asked and if you have too many bad verifications, ie you're trying to use a partial read, they might drop by to check out your reader... Two other problems, if it gets stolen, you're SOL. Second, the reader is likely to be expensive, making it hard to use this to allow purchase authorization at home. So your problem... The problem with this is that the validation server would have to know what the right answers are to all of the possible questions, and that creates a problem: either there would be waay too much data stored for each card, or there would only be a limited number of "questions" the server could ask. The answer, a limited number of questions. This would probably be fine tuned to balance out the replacement cost, anticipated number of validations during the lifetime of the thing etc. Seems like storage might be an issue though. As far as the half baked work around everyone else seems to be proposing, reading the article helps. The only one which actually might work, reproducing the resin using the paired laser/heat harden resin approach might actually work at some point. But it would require having the fob in the theif's possession for so long that the original would probably have noticed as missing, canceled the old one and gotten a new one by the time it was ready.