I work for a coupon company and we are on the Universal Code Council (UCC). One of the members of the UCC sits in the cube next to me. I can confirm; the first 5 digits are per manufacturer and are called a Product Family Code. What's worse, for easier accounting and administration, some manufacturers have more than one Product Family Code; for instance, I believe Nabisco has quite a few.
I agree. I think #3 is great for HP. But, I don't like the idea of replacing it every now and then. I just thought I'd point out that, economically speaking, of the 3 phase-transition technologies, this is in fact a pretty good one.
The problem with a liquid-gas phase transition is that the transition has to happen at just the right temperature. Water's gas phase transition is too high to be useful in cooling chips... most chips are rated right around 100C for max operating temp, active cooling that only works after the max operating temp is reached is in general a bad idea.
There are really only three ways to do phase transition cooling that make sense.
1. You can use water but lower the pressure in the gas portion of the system. This has the disadvantage of requiring low pressure fittings and whatnot, and being somewhat difficult to manufacture. Also, you'd probably need a compressor to maintain a pressure imbalance.
2. You can use Freon or whatever the replacement chemical is. This has the opposite problem of water; in order for the Freon to stay liquid, you have to have a high-pressure system. Compressor is mandatory. This would be an air conditioner with the cold coils directly on the die.
3. You can custom make a liquid coolant with a phase transition in a sane place. Like 50C or so. Rubbing Alcohol would work, but you have to be careful. Alcohol eats plastic and corrodes metals. Rubber too. So everything including your pump has to be made to withstand constant alcohol vapor. You could make a different liquid thats not so corrosive. In either case, your compressor/pump would be MUCH smaller than in the other cases.
Re:Wanna hurt the MPAA/RIAA'? End Media Addiction
on
Borrowing ROMs
·
· Score: 2
Here goes all of my karma:
Settlers of Catan rocks! I thought I was the only guy that knew it existed!!!
Similar story; in order to test security in an application I was auditing, one step of the signup process is that an account representative needs to verify that a particular person works for the client that (s)he said s(he) represents. So I picked a likely client (large client, no way the rep knows everyone at the company), and contacted the account rep's manager letting her know I was going to do it, but to act normally otherwise.
Signed up for an account with a real name from a real address in the companies home town. 15 minutes later the security director gets a call from their account rep that they believe someone maliciously tried to gain access.
Did quite a few people get concerned over this? Was there a risk of damaging their reputation with the client? Did it cost lots of time? Absolutely. But now the president of that company knows that his staff has been properly trained in security procedures, and he thought that was more valuable than the harm (potential and otherwise) that was caused.
Ummm... there's about three different ways to do it (well):
1. Map My Documents to a remote share on your file server, backup that drive, and setup policies so that data can't get saved locally.
2. Tell the users that local data may be deleted at any time, and they're supposed to use the space you setup at \\file_server\home_directories. Have a nice long meeting drilling in the point, then remind them the day before (do it in the morning and right after lunch) that you'll be deleting local files overnight. Do a fresh install of your OS of choice, configure it to access the home directory of whatever user is logged in, then ghost it over. This has the advantage that any computer is sufficient for the user's needs, as every computer is identically configured.
3. This is the least intrusive, and thus probably best. But I didn't see it mentioned yet (browsing at +1), so thought I'd say it. Most Windows machines share their drives out by default under administrative names... \\computer_name\c$, \\computer_name\d$, etc. You may need to configure Win98 to share out its drives, but thats trivial. Have a network backup server that just backs up the contents of each of these shares in turn. The users never even need to know that you did it, until something bad happens
Actually, there are a certain number of redundencies... not sure if its cause they're built in, or there are just two or three different ways to make a certain protein...
Also, IIRC half the codespace is taken up by the left/right dichotomy... Each amino acid can be left/right, but for some reason all of them are right-handed. The tRNA that decodes the DNA must convert all of the code sequences to right-handedness as it builds amino acids. That way, you can decode EITHER half of a DNA strand and get the same results...
Actually, using your monopoly position to undercut the competition by dumping your product on the market well below the market price is illegal, as I recall.
I think its a little touchy... in the XBox case, they don't have a monopoly in the market. But, as I learned not long after I posted this comment, product dumping CAN be illegal, but its also hard to prove in court. One of the reasons the states went after the intrinsic bundling of IE with Windows as opposed to product dumping in the recent court case.
I have a quick comment amount Microsoft giving away stuff for cheap: Purposefully selling a product at a loss is not illegal. Using a monopoly to shut out other markets IS... so, selling XBoxes at a low price (or even free) is not illegal, neither is giving away Internet Explorer.
Now, if Microsoft were selling XBoxes or giving them away, and they were engineered so that a TV that has an XBox connected to it cannot also have a PS2 or Gamecube or console system Y connected to it at the same time, that would be illegal. Or if Microsoft made, say, televisions instead of Operating Systems and engineered their TVs to only work (well) with Xboxes, that also would be illegal.
That said, I think that it definitely counts as unfair market practices, just not under the scope of current law. But I don't think government subsidies of the gaming industry is the answer...
I don't know what the answer is, but I don't think thats it. Maybe changing the law so that using money gained from a monopoly to establish a monopoly in another market is illegal... something along those lines...
I used to think much the same... then I tried to write my own chess-playing program. It ended up being surprisingly difficult!
First off, no computer, anywhere, has enough storage capacity. I heard elsewhere in this thread that there are more legal positions in chess that can be reached by a legal sequence of moves than there are atoms in the universe. I don't know about that; what I do know is that once you get into the fifth or sixth move, you start running into volatile memory limits. 20 moves or so is the limit for a standard desktop utilizing all of its available hard drive to store the table. Before that point, time to compute the next level on the tree starts approaching infinity.
The second point is, that at no time can a computer guarantee a win. That's called solving for mate, that is, showing that no matter what moves the opponent has, there is a corresponding move that you can make that guarantees you will eventually checkmate the opponent. I've seen a solve for mate a turn ahead, and even two turns ahead. Never farther. Chess is such a robust and open game in terms of possible moves that there's no way to force the opposing player onto a branch of a move tree that absolutely guarantees a win.
Now, what my program (and I imagine many others) did was find the move that had the most chance that the computer would eventually be able to solve for mate, while also avoiding branches where the player would be able to solve for mate. You could change scoring preferences to make it more agressive or more defensive, but that was about it.
I imagine that the best way for a computer to learn to play chess would be a neural network based solution. The above situation was deterministic; a good player (myself and my brother who had no knowledge of the coding) could predict after the 10th move or so what the computer would do, and formulate a strategy specifically meant to trap the computer on a good-looking tree where it is possible to solve for mate relatively easily... a neural network would be somewhat non-deterministic and would specifically be able to deal with that contingency much better.
If you had read the article, all four of the reviewed software packages use a non-MPAA approved css decoder. And are free. So, what exactly is it that you are griping about?
You are absolutely correct. The Supreme court has long maintained that the law cannot concern itself with the content of speech. Only whether or not it counts as speech.
This is, obviously, to protect those who publish such works. If a work of some sort (be it a public speech, a book, a newspaper, or a painting) qualifies as the first ammendment version of speech, then the law cannot pay any attention to the content of that work.
All that said, I would support some constitutional way for me, as an avid sci-fi reader and supporter of NASA, to flag my tax dollars as going to NASA. Would I donate directly to NASA? No, that's ludicrous... that's why I pay my taxes.
On the other hand, if there were any decent privately funded space programs in the world, I would purchase stock in a heartbeat. The difference being, I already own a partial share of NASA, by virtue of being an American citizen, and will reap whatever benefits come from the space program. If it were a privately funded endeavor, I would gladly pay to help fund that endeavor, and to be able to reap the rewards later.
I am moderately experienced in this arena; my step-mother is blind, and I am currently helping a friend who is losing her vision.
My step-mother, who is now in her fifties, has been blind since she was 9. She was given full classes in braille and has used it most of her life. However, she does not prefer it. I've seen her read braille; its painfully slow. She can usually do better with her text-to-speech program than with braille. I don't know many blind people that can read braille at anything like elementary reading speeds. Just my $.02, but it's intrinsically faster to scan a line of text with your eye (that doesn't have to move; just pivot) than to run your fingers over it and decode it one character at a time. My point? It's a mistake to think of braille as writing for the blind; its an alternative for those situations where a blind person has to read or convey information in a permanent, written manner, but don't fool yourself into thinking that it's a drop-in replacement.
As far as writing, I have never met a blind person that writes in braille. I have seen typewriters whose keys are both printed and in braille, and the typewriter embosses the paper in braille instead of printing in ink, but that's it. Most blind people I've met who are lucky enough to have family that will help them, actually write in english and have their SO or other family member read their writing for them.
Re:The best he can build is a disintegration chamb
on
Time Travel
·
· Score: 2
Actually, he can send things into the future all the time... it's just into the past. The real question here, is, without actually seeing evidence of time travel, how does one tell if we're in the "real", backwards-time-travellerless, universe or a parallel one, without invoking Occam's Razor, which is not a proof but a maxim.
What's more, one could presume that in every parallel universe into which a time traveller travels, more time travellers travel into the past, because they have concrete proof that it is possible, and access to the technology. Might there not be some bubble universe where man DID live alongside the dinosaurs, simply because enough people travelled into the past while spawning off parallel universes...
Bernstein had a fundamental problem with his proposal... basically the proposal is a diagram of how to make a large number of cheap (low memory, high speed) processors work together on a decryption. He does have some novel hardware devices that can speed up the process, but his supposed speed improvements all rely on one critical assumption:
That these processors are connected in a network that is zero-latency and infinite bandwidth.
The only way to truly provide something like that would be a network whose total bandwidth is greater than the combined clock speed of all the processors. That means that if you have 2^n processors, you need n network connections for each processor, in a large parallel array, commonly known in parallel computing as a hypercube. Let's say you build a machine of 8 100MHz 8-bit processors. Each processor gets 3 connections, all of which are 1Gbps. The total number of these connections are equal to the number of edges in a cube: 12, because that's how these processors are connected. Each processor is at a vertex and each edge is a connection. Add 8 more processors, and you get 32 connections.
For any machine sufficiently large enough to break a 1024-bit key, the cost would currently be greater than the entire economic output of earth for years. No doubt as high-speed networking technologies continue to be reduced in cost, this will change, but right now, Bernstein's proposal is unimplementable.
Honestly, I'm kinda hoping for the second option. I would never use this silliness to encode anything more secure than a shopping list, but as a security consultant, I can't help but thinking that one's failure is another's boon, and all that...
First, I'd like to point out that I said most. This certainly doesn't apply to all. But, every symmetric encryption algorithm I've ever seen works like this: it takes a random-looking number and XOR's it with the plaintext to generate the cipher text. If you don't have a random number, you use your non-random number as a seed in a random-number generator to make it random.
But, if one were to use keylength as a measurement of the security of an algorithm; well, in symmetric encryption, if you get a certain number of consecutive bits correct, the result looks less random. One can assume therefore that those bits are correct, and start focusing on the rest of the bits.
While this has been an excellent read, and I thank you greatly for providing it, I would like to make a couple observations:
Firstly, the random function he produces depends on a 200-page equation using variables for exponents. The equation is deterministic, but random. If you provide the same exponents, then the halting probability will be the same every time. However, any particular value of Omega, when considered independently, appears to be mathematically random. This is good stuff, don't get me wrong. Provably random-looking numbers. VERY good stuff. But, for the purposes of cryptography, not useful, because in order to prove it secure, it has to be open; therefore, finding out how the number was generated is pretty simple, and your entropy is reduced back to the number of bits in the variables used as exponents, because that's your search space.
This WOULD be a great way of further increasing the entropy in a hardware random number generator, however.
Ummm... no mathematical function is random. While you are correct, in that, given a random bitstream, every permutation you try will look like the correct decryption (and a large portion of them will generate something that looks like english, but isn't the original message)... the problem is the bit-choosing algorithm. You could take every bit, but then someone with access to the signal would be able to easily decode. You could take every other bit, but that isn't much more difficult.
You could use a large polynomial function; in that case, the amount of entropy in your encryption will be equal to the amount of entropy in your polynomials; if you can describe them in 128-bits, then that's the amount of entropy in your cipher.
Of course, all of this assumes that Eve intercepted the satelite signal at the same time.
Security is only potentially higher IF the one-time pad is communicated outside of electronic channels (ex: secured courier delivering pad directly into electronic safe), which is not what they're doing.
But, you're absolutely right about the above method (and any other secure one) being impractical in the real world; its generally only used for the most secret of secrets...
1024 bit, while not unbreakable, is still unbreakable in the lifetime of the universe. I have no doubt methodologies and processes will be developped in the future that will change this, but as of right now, for all intents and purposes, it's unbreakable
Secondly, many parts of quantum mechanical behaviour *are* random, especially at macroscopic scales. For example, when a particular radioactive isotope chooses to decay is completely random; I've seen military random number generators that depend on this or similar effects to create truly random number.
But, no purely software random number generator will ever even come close to approaching randomness.
Once the server is set up with E2Sec, anyone who logs on through a Web browser or Internet link will automatically be given an encrypted connection. A small 4- to 10-kilobit file, a bit like a Web cookie, is loaded into the client computer's memory. The file contains a program to generate random encryption keys, so that the keys themselves don't have to be sent over the network connection. The program is so tiny that even the low-powered processors in a cellphone can run it with ease, Mr. Kassam said.
This is really unbreakable. Unless you happen to intercept this program. Which wouldn't be that hard, and it may in fact be the same program for every client. And, they're touting this for wireless communications.
I found this next part interesting:
The client generates a series of random numbers to use as an encryption key. This is number is exchanged with the server through a secure process known only to Prescient, the server uses it to encrypt any information it sends back to the client, and then the key is destroyed and a new one is created. This process is repeated every time information is exchanged between the client and the server, making it virtually impossible for outsiders to decrypt the information.
It's a well established fact that non-open, secure processes are not secure. Cryptography is difficult, folks. The only way to even come close to proving that a particular process is secure is by exposing it to the scrutiny of the entire global community. Even then, its a case of proving that something is NOT true, which in this case involves incredibly complex mathematics that don't work for half of the proposed protocols out there; for instance, for a particular protocol to be 'provably' secure, it has to be time reversible (that is, if you apply any one step in reverse, the encryption key and cipher text each go back to their state before that step)
"We're 100-per-cent confident in our technology," Mr. Kassam said. "To give an idea of how difficult this is to crack, many organizations consider 128-bit encryption, which has a [cryptography level] of two to the power of 128, to be very secure. With e2Sec, we're talking about encryption in excess of 5,000 bits, and as much as two to the power of 10,000."
Ummmm... comparing asymmetric encryption to symmetric encryption (of which a one-time pad is a subset) with key-lengths is like comparing apples to oranges. In asymmetric encryption, your security is in your keyspace... every bit doubles the time to search the keyspace. In symmetric encryption, security is all about the keys; symmetric encryption is so easy to do that you can try millions of keys a second, as opposed to thousands or hundreds, so you HAVE to have a big keyspace. But, most symmetric encryption algorithms allow you to get it partly right; if the key is partly right, you get a partly decoded message, so the search algorithm is linear instead of exponential.
Slashdot Mirror
I work for a coupon company and we are on the Universal Code Council (UCC). One of the members of the UCC sits in the cube next to me. I can confirm; the first 5 digits are per manufacturer and are called a Product Family Code. What's worse, for easier accounting and administration, some manufacturers have more than one Product Family Code; for instance, I believe Nabisco has quite a few.
I agree. I think #3 is great for HP. But, I don't like the idea of replacing it every now and then. I just thought I'd point out that, economically speaking, of the 3 phase-transition technologies, this is in fact a pretty good one.
The problem with a liquid-gas phase transition is that the transition has to happen at just the right temperature. Water's gas phase transition is too high to be useful in cooling chips... most chips are rated right around 100C for max operating temp, active cooling that only works after the max operating temp is reached is in general a bad idea.
There are really only three ways to do phase transition cooling that make sense.
1. You can use water but lower the pressure in the gas portion of the system. This has the disadvantage of requiring low pressure fittings and whatnot, and being somewhat difficult to manufacture. Also, you'd probably need a compressor to maintain a pressure imbalance.
2. You can use Freon or whatever the replacement chemical is. This has the opposite problem of water; in order for the Freon to stay liquid, you have to have a high-pressure system. Compressor is mandatory. This would be an air conditioner with the cold coils directly on the die.
3. You can custom make a liquid coolant with a phase transition in a sane place. Like 50C or so. Rubbing Alcohol would work, but you have to be careful. Alcohol eats plastic and corrodes metals. Rubber too. So everything including your pump has to be made to withstand constant alcohol vapor. You could make a different liquid thats not so corrosive. In either case, your compressor/pump would be MUCH smaller than in the other cases.
Here goes all of my karma:
Settlers of Catan rocks! I thought I was the only guy that knew it existed!!!
Check it out here
Actually, write-only media is incredibly useful... look at the widespread implementation of /dev/null
Similar story; in order to test security in an application I was auditing, one step of the signup process is that an account representative needs to verify that a particular person works for the client that (s)he said s(he) represents. So I picked a likely client (large client, no way the rep knows everyone at the company), and contacted the account rep's manager letting her know I was going to do it, but to act normally otherwise.
Signed up for an account with a real name from a real address in the companies home town. 15 minutes later the security director gets a call from their account rep that they believe someone maliciously tried to gain access.
Did quite a few people get concerned over this? Was there a risk of damaging their reputation with the client? Did it cost lots of time? Absolutely. But now the president of that company knows that his staff has been properly trained in security procedures, and he thought that was more valuable than the harm (potential and otherwise) that was caused.
Ummm... there's about three different ways to do it (well):
1. Map My Documents to a remote share on your file server, backup that drive, and setup policies so that data can't get saved locally.
2. Tell the users that local data may be deleted at any time, and they're supposed to use the space you setup at \\file_server\home_directories. Have a nice long meeting drilling in the point, then remind them the day before (do it in the morning and right after lunch) that you'll be deleting local files overnight. Do a fresh install of your OS of choice, configure it to access the home directory of whatever user is logged in, then ghost it over. This has the advantage that any computer is sufficient for the user's needs, as every computer is identically configured.
3. This is the least intrusive, and thus probably best. But I didn't see it mentioned yet (browsing at +1), so thought I'd say it. Most Windows machines share their drives out by default under administrative names... \\computer_name\c$, \\computer_name\d$, etc. You may need to configure Win98 to share out its drives, but thats trivial. Have a network backup server that just backs up the contents of each of these shares in turn. The users never even need to know that you did it, until something bad happens
Actually, there are a certain number of redundencies... not sure if its cause they're built in, or there are just two or three different ways to make a certain protein...
Also, IIRC half the codespace is taken up by the left/right dichotomy... Each amino acid can be left/right, but for some reason all of them are right-handed. The tRNA that decodes the DNA must convert all of the code sequences to right-handedness as it builds amino acids. That way, you can decode EITHER half of a DNA strand and get the same results...
Actually, using your monopoly position to undercut the competition by dumping your product on the market well below the market price is illegal, as I recall.
I think its a little touchy... in the XBox case, they don't have a monopoly in the market. But, as I learned not long after I posted this comment, product dumping CAN be illegal, but its also hard to prove in court. One of the reasons the states went after the intrinsic bundling of IE with Windows as opposed to product dumping in the recent court case.
I have a quick comment amount Microsoft giving away stuff for cheap: Purposefully selling a product at a loss is not illegal. Using a monopoly to shut out other markets IS... so, selling XBoxes at a low price (or even free) is not illegal, neither is giving away Internet Explorer.
Now, if Microsoft were selling XBoxes or giving them away, and they were engineered so that a TV that has an XBox connected to it cannot also have a PS2 or Gamecube or console system Y connected to it at the same time, that would be illegal. Or if Microsoft made, say, televisions instead of Operating Systems and engineered their TVs to only work (well) with Xboxes, that also would be illegal.
That said, I think that it definitely counts as unfair market practices, just not under the scope of current law. But I don't think government subsidies of the gaming industry is the answer...
I don't know what the answer is, but I don't think thats it. Maybe changing the law so that using money gained from a monopoly to establish a monopoly in another market is illegal... something along those lines...
I used to think much the same... then I tried to write my own chess-playing program. It ended up being surprisingly difficult!
First off, no computer, anywhere, has enough storage capacity. I heard elsewhere in this thread that there are more legal positions in chess that can be reached by a legal sequence of moves than there are atoms in the universe. I don't know about that; what I do know is that once you get into the fifth or sixth move, you start running into volatile memory limits. 20 moves or so is the limit for a standard desktop utilizing all of its available hard drive to store the table. Before that point, time to compute the next level on the tree starts approaching infinity.
The second point is, that at no time can a computer guarantee a win. That's called solving for mate, that is, showing that no matter what moves the opponent has, there is a corresponding move that you can make that guarantees you will eventually checkmate the opponent. I've seen a solve for mate a turn ahead, and even two turns ahead. Never farther. Chess is such a robust and open game in terms of possible moves that there's no way to force the opposing player onto a branch of a move tree that absolutely guarantees a win.
Now, what my program (and I imagine many others) did was find the move that had the most chance that the computer would eventually be able to solve for mate, while also avoiding branches where the player would be able to solve for mate. You could change scoring preferences to make it more agressive or more defensive, but that was about it.
I imagine that the best way for a computer to learn to play chess would be a neural network based solution. The above situation was deterministic; a good player (myself and my brother who had no knowledge of the coding) could predict after the 10th move or so what the computer would do, and formulate a strategy specifically meant to trap the computer on a good-looking tree where it is possible to solve for mate relatively easily... a neural network would be somewhat non-deterministic and would specifically be able to deal with that contingency much better.
If you had read the article, all four of the reviewed software packages use a non-MPAA approved css decoder. And are free. So, what exactly is it that you are griping about?
You are absolutely correct. The Supreme court has long maintained that the law cannot concern itself with the content of speech. Only whether or not it counts as speech.
This is, obviously, to protect those who publish such works. If a work of some sort (be it a public speech, a book, a newspaper, or a painting) qualifies as the first ammendment version of speech, then the law cannot pay any attention to the content of that work.
All that said, I would support some constitutional way for me, as an avid sci-fi reader and supporter of NASA, to flag my tax dollars as going to NASA. Would I donate directly to NASA? No, that's ludicrous... that's why I pay my taxes.
On the other hand, if there were any decent privately funded space programs in the world, I would purchase stock in a heartbeat. The difference being, I already own a partial share of NASA, by virtue of being an American citizen, and will reap whatever benefits come from the space program. If it were a privately funded endeavor, I would gladly pay to help fund that endeavor, and to be able to reap the rewards later.
I am moderately experienced in this arena; my step-mother is blind, and I am currently helping a friend who is losing her vision.
My step-mother, who is now in her fifties, has been blind since she was 9. She was given full classes in braille and has used it most of her life. However, she does not prefer it. I've seen her read braille; its painfully slow. She can usually do better with her text-to-speech program than with braille. I don't know many blind people that can read braille at anything like elementary reading speeds. Just my $.02, but it's intrinsically faster to scan a line of text with your eye (that doesn't have to move; just pivot) than to run your fingers over it and decode it one character at a time. My point? It's a mistake to think of braille as writing for the blind; its an alternative for those situations where a blind person has to read or convey information in a permanent, written manner, but don't fool yourself into thinking that it's a drop-in replacement.
As far as writing, I have never met a blind person that writes in braille. I have seen typewriters whose keys are both printed and in braille, and the typewriter embosses the paper in braille instead of printing in ink, but that's it. Most blind people I've met who are lucky enough to have family that will help them, actually write in english and have their SO or other family member read their writing for them.
Actually, he can send things into the future all the time... it's just into the past. The real question here, is, without actually seeing evidence of time travel, how does one tell if we're in the "real", backwards-time-travellerless, universe or a parallel one, without invoking Occam's Razor, which is not a proof but a maxim.
What's more, one could presume that in every parallel universe into which a time traveller travels, more time travellers travel into the past, because they have concrete proof that it is possible, and access to the technology. Might there not be some bubble universe where man DID live alongside the dinosaurs, simply because enough people travelled into the past while spawning off parallel universes...
Bernstein had a fundamental problem with his proposal... basically the proposal is a diagram of how to make a large number of cheap (low memory, high speed) processors work together on a decryption. He does have some novel hardware devices that can speed up the process, but his supposed speed improvements all rely on one critical assumption:
That these processors are connected in a network that is zero-latency and infinite bandwidth.
The only way to truly provide something like that would be a network whose total bandwidth is greater than the combined clock speed of all the processors. That means that if you have 2^n processors, you need n network connections for each processor, in a large parallel array, commonly known in parallel computing as a hypercube. Let's say you build a machine of 8 100MHz 8-bit processors. Each processor gets 3 connections, all of which are 1Gbps. The total number of these connections are equal to the number of edges in a cube: 12, because that's how these processors are connected. Each processor is at a vertex and each edge is a connection. Add 8 more processors, and you get 32 connections.
For any machine sufficiently large enough to break a 1024-bit key, the cost would currently be greater than the entire economic output of earth for years. No doubt as high-speed networking technologies continue to be reduced in cost, this will change, but right now, Bernstein's proposal is unimplementable.
Honestly, I'm kinda hoping for the second option. I would never use this silliness to encode anything more secure than a shopping list, but as a security consultant, I can't help but thinking that one's failure is another's boon, and all that...
Alright, I was reaching a little on the last paragraph; but the doubling-your-keyspace argument was right from Applied Cryptography
First, I'd like to point out that I said most. This certainly doesn't apply to all. But, every symmetric encryption algorithm I've ever seen works like this: it takes a random-looking number and XOR's it with the plaintext to generate the cipher text. If you don't have a random number, you use your non-random number as a seed in a random-number generator to make it random.
But, if one were to use keylength as a measurement of the security of an algorithm; well, in symmetric encryption, if you get a certain number of consecutive bits correct, the result looks less random. One can assume therefore that those bits are correct, and start focusing on the rest of the bits.
While this has been an excellent read, and I thank you greatly for providing it, I would like to make a couple observations:
Firstly, the random function he produces depends on a 200-page equation using variables for exponents. The equation is deterministic, but random. If you provide the same exponents, then the halting probability will be the same every time. However, any particular value of Omega, when considered independently, appears to be mathematically random. This is good stuff, don't get me wrong. Provably random-looking numbers. VERY good stuff. But, for the purposes of cryptography, not useful, because in order to prove it secure, it has to be open; therefore, finding out how the number was generated is pretty simple, and your entropy is reduced back to the number of bits in the variables used as exponents, because that's your search space.
This WOULD be a great way of further increasing the entropy in a hardware random number generator, however.
Ummm... no mathematical function is random. While you are correct, in that, given a random bitstream, every permutation you try will look like the correct decryption (and a large portion of them will generate something that looks like english, but isn't the original message)... the problem is the bit-choosing algorithm. You could take every bit, but then someone with access to the signal would be able to easily decode. You could take every other bit, but that isn't much more difficult.
You could use a large polynomial function; in that case, the amount of entropy in your encryption will be equal to the amount of entropy in your polynomials; if you can describe them in 128-bits, then that's the amount of entropy in your cipher.
Of course, all of this assumes that Eve intercepted the satelite signal at the same time.
Security is only potentially higher IF the one-time pad is communicated outside of electronic channels (ex: secured courier delivering pad directly into electronic safe), which is not what they're doing.
But, you're absolutely right about the above method (and any other secure one) being impractical in the real world; its generally only used for the most secret of secrets...
I have two things to say:
1024 bit, while not unbreakable, is still unbreakable in the lifetime of the universe. I have no doubt methodologies and processes will be developped in the future that will change this, but as of right now, for all intents and purposes, it's unbreakable
Secondly, many parts of quantum mechanical behaviour *are* random, especially at macroscopic scales. For example, when a particular radioactive isotope chooses to decay is completely random; I've seen military random number generators that depend on this or similar effects to create truly random number.
But, no purely software random number generator will ever even come close to approaching randomness.
From the article:
Once the server is set up with E2Sec, anyone who logs on through a Web browser or Internet link will automatically be given an encrypted connection. A small 4- to 10-kilobit file, a bit like a Web cookie, is loaded into the client computer's memory. The file contains a program to generate random encryption keys, so that the keys themselves don't have to be sent over the network connection. The program is so tiny that even the low-powered processors in a cellphone can run it with ease, Mr. Kassam said.
This is really unbreakable. Unless you happen to intercept this program. Which wouldn't be that hard, and it may in fact be the same program for every client. And, they're touting this for wireless communications.
I found this next part interesting:
The client generates a series of random numbers to use as an encryption key. This is number is exchanged with the server through a secure process known only to Prescient, the server uses it to encrypt any information it sends back to the client, and then the key is destroyed and a new one is created. This process is repeated every time information is exchanged between the client and the server, making it virtually impossible for outsiders to decrypt the information.
It's a well established fact that non-open, secure processes are not secure. Cryptography is difficult, folks. The only way to even come close to proving that a particular process is secure is by exposing it to the scrutiny of the entire global community. Even then, its a case of proving that something is NOT true, which in this case involves incredibly complex mathematics that don't work for half of the proposed protocols out there; for instance, for a particular protocol to be 'provably' secure, it has to be time reversible (that is, if you apply any one step in reverse, the encryption key and cipher text each go back to their state before that step)
"We're 100-per-cent confident in our technology," Mr. Kassam said. "To give an idea of how difficult this is to crack, many organizations consider 128-bit encryption, which has a [cryptography level] of two to the power of 128, to be very secure. With e2Sec, we're talking about encryption in excess of 5,000 bits, and as much as two to the power of 10,000."
Ummmm... comparing asymmetric encryption to symmetric encryption (of which a one-time pad is a subset) with key-lengths is like comparing apples to oranges. In asymmetric encryption, your security is in your keyspace... every bit doubles the time to search the keyspace. In symmetric encryption, security is all about the keys; symmetric encryption is so easy to do that you can try millions of keys a second, as opposed to thousands or hundreds, so you HAVE to have a big keyspace. But, most symmetric encryption algorithms allow you to get it partly right; if the key is partly right, you get a partly decoded message, so the search algorithm is linear instead of exponential.