Both the summary and the article are a little light on details, however the article mentions replacing, (or extending) the arithmetic (lossless) encoder - i.e. Huffman - used within the JPEG and H264 standards.
This would result in a lossless reduction in size of those files.
Again, short on details. Any size reduction claims are sorta hand wavy without more details.
But I'd think the loss-less label (or bit-exact) are ok in this context. Loss less from Jpeg -> DropJpeg.
Get any of the cheap Kyocera. They're indestructible, and inexpensive (~$20 new). My S2100 last easily 2 weeks per charge. Probably closer to 3 with a new battery. I like the flip better than the candy-bar style as I'm abusive with electronics that I carry around. I think the flip protects the screen more...
No keyboard, no touch. Yes, I need to hit '1'-'1'-'1' for C, etc. when TXTing. But I hardly ever send TXTs, so it works for me. Calls are fine.
Actually, the system is effectively sexing people. And the system's determination of sex, disagreed with the manual input - hence the flag for screening.
I've learned something new today. The TSA has machines that are discriminating passengers based on gender:)
We still use scopes quite a lot, and I don't see that changing. Logic analyzers however - they sit gathering dust. With more embedded devices offering similar functions, then LA's just been replaced in our house. FPGAs have "scope" tools (which are really LA's with more limited depth). Better trace functions on debuggers. All these work fine for "digital" type problems. Outside this debug - you're firmly in in analog land, and a good scope is a must.
The articles right in that one of the first things to be hit will be kid's toys. Forget printing cars and clothes like this thread's talking about.
The first big lawsuit:
Legos.
No complex shapes, plastic that fits what 3D printing can do. A deep pocket industry, where they'll feel the effects quickly. Teens that can probably come up with the basic shapes with trial and error in just a few hours.
Yeah, where's the popcorn? This is going to be show...
The article doesn't say anything about returning the mined material to earth. I gotta think the intrinsic value of any mined material is worth much more in orbit rather than back down on Earth. Getting raw materials up into orbit is very expensive.
Back in the late 80s, Georgia Tech would have any incoming freshmen with lower high school GPAs start in the Summer quarter. This was under the auspices of giving those who were struggling, a bit more time to adjust to college curriculum before the incoming fall crush.
The interesting "side effect" was that the GPA of incoming Fall freshmen was thus higher, and the university had no trouble repeating that fact.
My point is that it doesn't have to be removed -- just corrupted.
Same thing => remove = corrupt in this case.
I contend that with publically knowable algorithms, but secret keys, you can reliably (read reliable here as redundant, with a checksum ) hide data in a mp3. At least to the point where "corrupting" it to sufficiently remove the watermark also corrupts the usefulness of the mp3 at at all. Plus the user could never be "sure" that it's completely removed.
This is why the military uses spread spectrum. The enemy can know that there's information being transmitted. But without the secret key, he cannot decipher it, or replace it with misinformation. He can only "drown it out" by filling the spectrum with noise. The military does go the extra step in "obscuring" the algorithm - but that's icing on the cake.
So, without knowing anything about the supersecret watermarking scheme, you can still take two (or more) otherwise identical watermarked files and compare them. Any differences at all will be the difference between the watermarks. As an attacker, if you can find those differences you can corrupt them. You don't need to "crack the code" to figure out the exact value of my watermark because it's irrelevant to you. You don't care if the hidden value is 17 or 42 -- it is simply enough for you to damage the watermark beyond my ability to decode it. Once you've done that, the file is untraceable and you can safely make it public without fear of retribution from me.
Google "Spread Spectrum". The same techniques here can be used for watermarking/signature (They are variants of the same thing in my mind). A "watermark" is just a signature that's very difficult to remove. The watermark can be hidden in the frequency bands just like a spread spectrum signal. Well, the spectrum of mp3 is not like that of a wireless signal, but those differences can be accounted for.
This all works so well because the bandwidth of the watermark is soo small - there's soo many places to hide the data (with plenty of redundancy!) Pick one (via a secret key), and hide the data there. In all the other places modulate the real signal with a little random noise.
Sure many people can diff there files, and see that EVERY mp3 is different in some small way - now how would one scrub it - keep zeroing out ALL differences? You end up degrading the song until it's useless. And again - here's the key - where's the incentive to remove it?
Watermarking is 100% "security through obscurity."
Not correct. DRM is 100% "security through obscurity". The way I understand that this watermarking technology is to be used, it is a secure solution. The players would not have the means of detecting the watermark - if they did, then it would be just some other obscurity that is broken weeks after releasing.
A lot of the threads here are confusing DRM and watermarking. Anything with DRM is just security through obscurity. DRM schemes will invariably be broken, as the players have the secret key.
With watermarking, however, the player does not have the secret key.
This type of watermarking is the only viable long term solution, and one that I think will eventually be quite universal, and accepted. And yes, it could (and should!) be open sourced. The algorithm could be completely open - with just the "key" being kept secret.
Think about it, you need to hide what - let's conservatively say 512 bits - 64 bytes - into files that
are normally multi-megabytes in length? I have no experience in this area, but off the top of my head I can think of 3-4 ways of easily doing this, with redundancy. This is an arms race which the studios can win. They can't with DRM.
And for Joe Sixpack - he's bought some mp3 online - they contain some bit of hidden watermark in the file that indicated "Joe Sixpack bought this disk". He never knows this nor cares - he can download the mp3 to his iPod. Save it to a CD to play in the car, etc. Why should he even bother with any stripper program on the net? He has no incentive. Most of the reasons for anti-DRM programs on the net was to allow LEGAL users to listen to their LEGALLY purchased music, which for whatever reason DRM had broken.
But, when said mp3 file starts showing up all over the net then the studios has some evidence that says Joe Sixpack is illegally sharing files.
And to be honest, this is ok with me. The studios aren't going to go after kids sharing files with friends (well, at least when they final understand that it's isn't working, and hurting business). They will (and should) go after major pirates on the net, and this is a viable tool they could use.
Incorrect. ADSL uses a serially concatenated code. The outter code is the well known Reed-Soloman. The inner code is a convolutional interleaver.
Both are optional, and parameters are negotiated during modem training.
Some comments were brought up during the latests standards definitions about adding turbo codes to the ADSL specs. However, that code complexity issue raises it's head, and those discussions usually get dropped quickly...
I disagree here on the viability of digital watermarks. Think about it. You need to hide, perhaps 10 characters - 80 bits inside a comparitively large multi-megabyte source.
Heck make it 160 bits. Of course, redudancy adds data, but still we're talking a lot of places to hide data.
Prof. Felton's studies simply show what happens when you try to innovate via litigation rather than in an open, peer-reviewed environment. They (studios) started to take the first step, and then didn't even finish it. Engineering's an iterative process. They didn't even complete one iteration before the sharks got involved and the whole thing was killed. Heck in my designs, I've got a rule - Designs NEVER work right the first time; if it appears too, you've missed something.
If we had a bunch of bright people working on a open watermarking standard I'm confident that they could come up with a viable solution.
And I don't think that this would need to be tied to any "serial-number-tracking" system at all. It would simply give law-enforcement, and content provider a method for tracking large scale violations.
I agree here for the most part with a slight twist.
Instead of the serial number being in plaintext, and checked for validity, it should be a watermark entirely invisible for the end-user. Neither the hardware nor user would know the watermark was there.
This has a few advantages:
For the content creators, only they'd have the keys to view to watermark, making it MUCH harder to erase/disable.
What the content providers would have is a tool to track down the source of gross violaters of copyright, while still allowing fair-use.
For the end users, no checking for serial number validity to disable product. This has been a fairly useless feature even for software as the serial is plaintext, it's easy to disable the check.
People, people, people. We should really all be supporting this technology. This really is the answer that we want.
Let's make a few assumptions.
1. Someone can make a non-trivially breakable watermark technology. One that stands up to peer review without threats of legal ramifications.
2. Content providers can then use this watermarking technology on a reasonably fine scale - probably not individually watermarking every CD, but perhaps broken down into regions. Digital downloads could be individually watermarked, given enough CPUs.
What would this do? It gives the content providers ammunition and evidence to go after the big time copyright violators. Those that are burning CD's and turning around and charging money. Granted, a lot of these folks are probably overseas...
It allows us to use our digital media as we see fit. We can listen to it on our PCs. Download it to our Rios. It still allows us to swap digital media among friends. Content providers aren't going to go after the small fry, there's no return on investment.
This allows us to say to our congressmen, "Yes we care about and value copyrights. But we also value fair use."
This is a happy medium ground.
And being the crazy optimist that I am, this is the way I see things eventually settling down. The question is will it settle down in 1-2 years, or 10-20?
Dropped/wrong bits can be easily accounted for with ECC - i.e. adding redundancy information to the backup. It's basic information theory that's used today on all hard drives. A "Capacity" limit is determined for the drive ( or tape for this case ) based on the bit-error-rate. A redudancy code is created based on the BER.
Bammo - fractionally less than 600 GB at $10 tape. I'd buy it in a heartbeat.
Actually, I think their math and/or description must be wrong. I can't believe the tape would hold 600 GB. The data rates themselves seem to be overwhelming - the thing would have to write to the tapes at 75 GB / 30 min =~ 42 MB/sec!!. Only todays top of the line disk drives can hit this rate.
I've used unix in my job for years. So I don't need a "Dummies" book which holds your hand through the command line. I'm comfortable with command lines, makefiles, shells etc... I've even hacked Makefiles from distributions to get them to compile on my system.
But, for sysadmin type stuff, I'm lost. I've always relied on someone else. I just type:
log_problem, wait an hour or so, and POOF magic, the phone rings and someone tells me the thing is fixed.
So, now I'm dabbling in linux at home, and find I'm quite lost in the sysadmin type stuff. NIS? NFS?/etc/fstab, init.rc? Can someone explain what BSD vs SYS5 is and why folks are so argumentative about it...
I can read man pages no problem. But if one can't remember the command or really has problems remember where a config files is kept, you're in trouble.
Can someone recommend a book that fits this hole?
Linux Essential Reference above looks like it might be close. Looking for other inputs though...
Slashdot Mirror
Both the summary and the article are a little light on details, however the article mentions replacing, (or extending) the arithmetic (lossless) encoder - i.e. Huffman - used within the JPEG and H264 standards.
This would result in a lossless reduction in size of those files.
Again, short on details. Any size reduction claims are sorta hand wavy without more details.
But I'd think the loss-less label (or bit-exact) are ok in this context. Loss less from Jpeg -> DropJpeg.
Get any of the cheap Kyocera. They're indestructible, and inexpensive (~$20 new). My S2100 last easily 2 weeks per charge. Probably closer to 3 with a new battery. I like the flip better than the candy-bar style as I'm abusive with electronics that I carry around. I think the flip protects the screen more...
No keyboard, no touch. Yes, I need to hit '1'-'1'-'1' for C, etc. when TXTing. But I hardly ever send TXTs, so it works for me. Calls are fine.
Actually, the system is effectively sexing people. And the system's determination of sex, disagreed with the manual input - hence the flag for screening.
I've learned something new today. The TSA has machines that are discriminating passengers based on gender :)
We still use scopes quite a lot, and I don't see that changing. Logic analyzers however - they sit gathering dust. With more embedded devices offering similar functions, then LA's just been replaced in our house. FPGAs have "scope" tools (which are really LA's with more limited depth). Better trace functions on debuggers. All these work fine for "digital" type problems. Outside this debug - you're firmly in in analog land, and a good scope is a must.
The articles right in that one of the first things to be hit will be kid's toys.
Forget printing cars and clothes like this thread's talking about.
The first big lawsuit:
Legos.
No complex shapes, plastic that fits what 3D printing can do. A deep pocket industry, where they'll feel the effects
quickly. Teens that can probably come up with the basic shapes with trial and error in just a few hours.
Yeah, where's the popcorn? This is going to be show...
The article doesn't say anything about returning the mined material to earth. I gotta think the intrinsic value of any mined material is worth much more in orbit rather than back down on Earth. Getting raw materials up into orbit is very expensive.
Back in the late 80s, Georgia Tech would have any incoming freshmen with lower high school GPAs start in the Summer quarter. This was under the auspices of giving those who were struggling, a bit more time to adjust to college curriculum before the incoming fall crush.
The interesting "side effect" was that the GPA of incoming Fall freshmen was thus higher, and the university had no trouble repeating that fact.
I contend that with publically knowable algorithms, but secret keys, you can reliably (read reliable here as redundant, with a checksum ) hide data in a mp3. At least to the point where "corrupting" it to sufficiently remove the watermark also corrupts the usefulness of the mp3 at at all. Plus the user could never be "sure" that it's completely removed.
This is why the military uses spread spectrum. The enemy can know that there's information being transmitted. But without the secret key, he cannot decipher it, or replace it with misinformation. He can only "drown it out" by filling the spectrum with noise. The military does go the extra step in "obscuring" the algorithm - but that's icing on the cake.
This all works so well because the bandwidth of the watermark is soo small - there's soo many places to hide the data (with plenty of redundancy!) Pick one (via a secret key), and hide the data there. In all the other places modulate the real signal with a little random noise.
Sure many people can diff there files, and see that EVERY mp3 is different in some small way - now how would one scrub it - keep zeroing out ALL differences? You end up degrading the song until it's useless. And again - here's the key - where's the incentive to remove it?
A lot of the threads here are confusing DRM and watermarking. Anything with DRM is just security through obscurity. DRM schemes will invariably be broken, as the players have the secret key.
With watermarking, however, the player does not have the secret key.
This type of watermarking is the only viable long term solution, and one that I think will eventually be quite universal, and accepted. And yes, it could (and should!) be open sourced. The algorithm could be completely open - with just the "key" being kept secret.
Think about it, you need to hide what - let's conservatively say 512 bits - 64 bytes - into files that are normally multi-megabytes in length? I have no experience in this area, but off the top of my head I can think of 3-4 ways of easily doing this, with redundancy. This is an arms race which the studios can win. They can't with DRM.
And for Joe Sixpack - he's bought some mp3 online - they contain some bit of hidden watermark in the file that indicated "Joe Sixpack bought this disk". He never knows this nor cares - he can download the mp3 to his iPod. Save it to a CD to play in the car, etc. Why should he even bother with any stripper program on the net? He has no incentive. Most of the reasons for anti-DRM programs on the net was to allow LEGAL users to listen to their LEGALLY purchased music, which for whatever reason DRM had broken.
But, when said mp3 file starts showing up all over the net then the studios has some evidence that says Joe Sixpack is illegally sharing files.
And to be honest, this is ok with me. The studios aren't going to go after kids sharing files with friends (well, at least when they final understand that it's isn't working, and hurting business). They will (and should) go after major pirates on the net, and this is a viable tool they could use.
Incorrect. ADSL uses a serially concatenated code. The outter code is the well known Reed-Soloman. The inner code is a convolutional interleaver.
Both are optional, and parameters are negotiated during modem training.
Some comments were brought up during the latests standards definitions about adding turbo codes to the ADSL specs. However, that code complexity issue raises it's head, and those discussions usually get dropped quickly...
...the RIAA doesn't enemies. I mean common, is there any sane person who doesn't think this is just plain silly...
We need more companies making statements like these.
Go SunComm...What else you got?
I disagree here on the viability of digital watermarks. Think about it. You need to hide, perhaps 10 characters - 80 bits inside a comparitively large multi-megabyte source.
Heck make it 160 bits. Of course, redudancy adds data, but still we're talking a lot of places to hide data.
Prof. Felton's studies simply show what happens when you try to innovate via litigation rather than in an open, peer-reviewed environment. They (studios) started to take the first step, and then didn't even finish it. Engineering's an iterative process. They didn't even complete one iteration before the sharks got involved and the whole thing was killed. Heck in my designs, I've got a rule - Designs NEVER work right the first time; if it appears too, you've missed something.
If we had a bunch of bright people working on a open watermarking standard I'm confident that they could come up with a viable solution.
And I don't think that this would need to be tied to any "serial-number-tracking" system at all. It would simply give law-enforcement, and content provider a method for tracking large scale violations.
I agree here for the most part with a slight twist.
Instead of the serial number being in plaintext, and checked for validity, it should be a watermark entirely invisible for the end-user. Neither the hardware nor user would know the watermark was there.
This has a few advantages:
For the content creators, only they'd have the keys to view to watermark, making it MUCH harder to erase/disable.
What the content providers would have is a tool to track down the source of gross violaters of copyright, while still allowing fair-use.
For the end users, no checking for serial number validity to disable product. This has been a fairly useless feature even for software as the serial is plaintext, it's easy to disable the check.
People, people, people. We should really all be supporting this technology. This really is the answer that we want.
Let's make a few assumptions.
1. Someone can make a non-trivially breakable watermark technology. One that stands up to peer review without threats of legal ramifications.
2. Content providers can then use this watermarking technology on a reasonably fine scale - probably not individually watermarking every CD, but perhaps broken down into regions. Digital downloads could be individually watermarked, given enough CPUs.
What would this do? It gives the content providers ammunition and evidence to go after the big time copyright violators. Those that are burning CD's and turning around and charging money. Granted, a lot of these folks are probably overseas...
It allows us to use our digital media as we see fit. We can listen to it on our PCs. Download it to our Rios. It still allows us to swap digital media among friends. Content providers aren't going to go after the small fry, there's no return on investment.
This allows us to say to our congressmen, "Yes we care about and value copyrights. But we also value fair use."
This is a happy medium ground.
And being the crazy optimist that I am, this is the way I see things eventually settling down. The question is will it settle down in 1-2 years, or 10-20?
... 10 points if you know what "TRON" stood for.
-- I don't remember if they explained it
in the movie or not.
... 1000 points if you ever used TRON to debug
your program.
Fun times. Now we have purty IDE's and debug
tools that a moron could use... In my day...
Dropped/wrong bits can be easily accounted for with ECC - i.e. adding redundancy information to the backup. It's basic information theory that's used today on all hard drives. A "Capacity" limit is determined for the drive ( or tape for this case ) based on the bit-error-rate. A redudancy code is created based on the BER.
Bammo - fractionally less than 600 GB at $10 tape. I'd buy it in a heartbeat.
Actually, I think their math and/or description must be wrong. I can't believe the tape would hold 600 GB. The data rates themselves seem to be overwhelming - the thing would have to write to the tapes at 75 GB / 30 min =~ 42 MB/sec!!. Only todays top of the line disk drives can hit this rate.
Somethings fuzzy...
I've used unix in my job for years. So I don't need a "Dummies" book which holds your hand through the command line. I'm comfortable with command lines, makefiles, shells etc... I've even hacked Makefiles from distributions to get them to compile on my system.
/etc/fstab, init.rc? Can someone explain what BSD vs SYS5 is and why folks are so argumentative about it...
But, for sysadmin type stuff, I'm lost. I've always relied on someone else. I just type:
log_problem, wait an hour or so, and POOF magic, the phone rings and someone tells me the thing is fixed.
So, now I'm dabbling in linux at home, and find I'm quite lost in the sysadmin type stuff. NIS? NFS?
I can read man pages no problem. But if one can't remember the command or really has problems remember where a config files is kept, you're in trouble.
Can someone recommend a book that fits this hole?
Linux Essential Reference above looks like it might be close. Looking for other inputs though...