You can take a published string and make it a reasonably secure passphrase by adding enough entropy to it, but you still have to remember the entropy that you've added. Why not just start with a diceware passphrase and memorize the entropy directly?
I think that's still a valid point. How well can you misremember a quote? What is the maximum hamming distance between the original quote and a passphrase that you can remember? If you can remember 64 or more bits of entropy to add to an existing quote, you might as well remember a shorter diceware passphrase with the same entropy.
Additionally; how secret is your choice of source material? You can only have a finite number of books in your house, and a larger but finite number within driving distance. The likelihood of you traveling far and wide to generate a passphrase is pretty low. Can you be sure that Echelon didn't record the text (or at least the URLs) it's seen you fetch over your Internet connection? Reducing the search space to only a few thousand sources makes the problem almost embarrassingly simple. Build a probabilistic model of your writing/typing and then use it to find the nearest likely passphrases generated by altering the substrings of sources to better fit your writing style. Most likely you don't choose truly randomly from a set of altered quotes; you look for things in the text that seem easy to remember if they are changed, or that trigger some other memory that makes it easier to remember the other changes. Humans are bad at generating truly random text.
A 40 TB corpus has approximately 4 x 10^16 substrings of less than 1K, or 4 x 10^15 if we assume strings start on word boundaries and an average word length of 10 or less. Even if you charitably assume the whole hash/decrypt/validate process can be done in 1 ms of compute time per candidate, thats 4 x 10^13 seconds (about 1M years) of compute time. Unless Amazon has drastically lowered their prices, I don't think you'll be getting that for $150...
And in ten years, it will only be a thousand years of compute time using your numbers. Ten years later, only a year. Do your secrets only need to last 20 years? I would also remind you of GPU password cracking where billions of cryptographic primitives per second per card is not unusual. Assuming a thousand salting/setup primitive operations per passphrase (a bit lower than my preference for LUKS, but I've seen plenty of software with fewer iterations) would drop your estimate to 1000 GPU-years for an attacker to mount an attack today. Just pay some down-and-out bitcoin miners to put their now-overpriced rigs to use.
I admit that I underestimated the dollar amount for a well designed cryptosystem with iterated salting during key setup for dramatic effect. Taking some basic numbers from the bitcoin folks, it looks like a GPU cryptographic primitive (sha256 or ripemd160) costs around 1e-6 Joules. 4e16 substrings times 1000 primitives costs 4e13 Joules, or about 11 megawatthours, or about <pinky>1 million dollars</pinky>. However, there are probably many redundancies in the 40 TB database and it could be ordered by those redundancies to search through the more common space of text with shorter substrings first making it more likely to find weak passphrases quickly. For a weak cryptosystem where a single cryptographic operation suffices to test a passphrase I was only an order of magnitude off.
What is the economic difference between an illegal alien invader and you?
Pays taxes? No difference.
Produces goods and services? No difference.
Consumes goods and services? No difference.
Possesses a particular piece of paper issued by the government at public expense? BINGO!
It's obviously foolish to use public text verbatim as a key. Common Crawl has a 40 TB dataset that costs approximately $150 to MapReduce on EC2. Any key that happens to be a (reasonably short, say under 1KB) substring of that data costs $150 to break. Any key within a short hamming distance of a substring in that database costs roughly 2^hamming_distance more to break; two changed bytes is only worth $600. I imagine that large organizations who care have much larger databases including the text of most published books. It's such an obvious idea and until you realize that attackers have access to all the public source data that you do it sounds like a good idea to just pick a random string from a book to use as a passphrase. Don't kid yourself; no matter how obscure or unpopular a song is there will be lyrics for it somewhere on the Internet, not to mention in published books.
You can take a published string and make it a reasonably secure passphrase by adding enough entropy to it, but you still have to remember the entropy that you've added. Why not just start with a diceware passphrase and memorize the entropy directly?
The problem with hyping a test of an emergency system is that its effectiveness drops nearly to zero around the time of the test. Even testing an alert system too often reduces its effectiveness. For a national alert system (and the September 11, 2001 attacks didn't even warrant an alert, apparently) it will probably end up being roughly useless. A propaganda tool at best.
A central authority doesn't need to be a fragile forest of fully trusted CAs like we have now. A much better solution would be for clients to have marginal trust in any individual x509 signature of a certificate, requiring at least N distinct signatures to validate the certificate, where N is great enough to significantly reduce the threat of enough compromised CAs signing an attacker's certificate to make it trusted. Inherited trust from an intermediate CA would only carry a portion of the trust placed in the root certificate of the chain so that each of the N signatures would have to come from independent organizations.
Basically, build a PGP-like web of trust out of the current CAs. It's even in the root CA's economic best interest because they'll sell N times as many certificates.
I think they are swamped with UNqualified candidates. Recruiting gives them a list of at least potentially qualified candidates. If their standard interview process has a false positive rate of 1% and they have thousands of candidates of whom 99% are unqualified, then 50% of their hires will be qualified. If their recruiters give them a list of candidates of whom 25% are qualified, their odds of a qualified hire go up to about 97%.
People who have only ever used a single programming language and can't even understand another language are not programmers. I'm not sure they're even human. Same goes for application usage.
Just to be clear, I both believe in the existence of God (every possible god, in fact), and the non-existence of god at the same time. I exist in multiple universes, some with gods and some without, but enough of those universes are similar enough that I am unable to determine with sufficient confidence which universe this instance of myself is actually in.
Does this make me a theist, an atheist, or perhaps a schrodingerist?
In the interest of more accurate communication I'd like to introduce some other labels that should be applied to individuals as appropriate.
aspontaneous generationist (one who does not believe in spontaneous generation)
aplanar Earther (one who does not believe in a flat Earth)
acentralist (one who does not believe that the Earth is the center of the universe)
atroll (one who does not believe in trolling Internet discussion forums)
adead (one who believes they are still living)
aphlogistonist (one who does not believe in the phlogiston theory)
aplatonist (one who does not believe in the platonic solid arrangement of the Heavenly Spheres)
ageometer (one who does not believe that Euclid's Geometry is axiomatic)
asarcasmite (one who does not believe in using sarcasm in an argument)
Feel free to chime in with other vitally important labels that we can add to this list.
git's fast, cryptographically validated, and widely used and tested. Put everything you care about into a git repository, then clone it wherever you have extra space for backups. Enable remote pulls to keep the backup versions up to date.
I use subversion for the same purpose now, but I think git is overall a much better choice and I'll probably switch eventually.
The OS doesn't have to trust the boot sector, it can verify the information using cryptographic signatures. If you treat the block devices as untrusted, you will do things like using checksums, CRCs, ECC, etc... don't forget that hard drives typically get 1 out of 10^15 blocks wrong in the normal course of doing business.
DRM is not the same as formal correctness. I can probabilistically trust public key methods to protect probabilistically correct code, or I can fully trust formally proven code. Given the track record of DRM and code-signing I would opt for formal correctness. Hardware reliability is another issue entirely. You can only have probabilistic trust in the hardware, but unless you are working with tamper-resistant systems you probably don't have to worry about direct attacks on hardware reliability and can use ECC and cryptography to ensure that your own hardware is operating correctly.
Since an untrusted file system driver would have to use the OS to read / write the block devices, you could present it with a read-only capability to the block device which contains the OS, stopping all modification. In the case of a need for an update, the OS could then use a read/write capability to do the update, and switch back when done. I'm sure someone with a good background in CS could figure out a more secure way of doing it.
How does the OS know which data blocks contain user data and which blocks contain OS data? How does the OS know the filesystem isn't lying about the data it's going to write? Suppose the OS has perfect control over when and where the file system can write data. The file system says "Time to install an OS upgrade", gets permission, and then writes a rootkit instead of valid data. The entire chain has to be trusted. "a more secure way of doing it" is to formally verify the data to be written to the OS disks before it's written, and then when booting formally verify it again before running it to be sure it wasn't modified.
Not if the memory management is working correctly... Windows NT 3.5 and earlier had video running outside the kernel... which is the right way to do things in terms of security.
It's not just the CPU that has access to RAM, any device on the PCI bus has access as well. You do have to trust your hardware, of course, but once the hardware is trusted you also need to trust the driver software to give the correct addresses to the hardware for DMA. The hardware has no idea what sort of protection the OS is using or how it has organized memory, it just reads and writes to specific addresses in RAM. There were (are) numerous X11 privilege escalations based on this fact since OpenGL programs can talk directly to the hardware via direct rendering.
Drivers don't have to run in kernel mode to be efficient, and the DMA doesn't have to be set up directly by the drivers. The MMU and the control over input parameters to the DMA/Interrupt subsystems should suffice here as well.
Sometimes the DMA parameters are sent directly to the hardware, especially with video cards. DMA scatter/gather is also used by many disk and network controllers. The particular interface used to tell the hardware which addresses to use is not standardized to the point that every possible hardware driver could use a simple, secure operating system service to pass those DMA parameters to the hardware.
The primary reason we even worry about root kits is that Operating Systems haven't been designed to work in a world of untrusted code. Changing that one aspect of things, doing a hell of a lot of coding to build a capability based OS, provides an environment in which it is very unlikely that any rootkit would get the opportunity to be installed.
So you have a perfect capability operating system and it will never be infected by a rootkit and you have fine-grained control over network and file access. You have a secret file on your computer that you would like to process, the result of the proc
Maybe you underestimate the resources of these cartels. They are almost certainly capable of assassinating the Mexican president and any other members of government, and I would not be surprised if they could assassinate any political leader on Earth. It's not hard. Get your hands on an air-to-surface missile and use one of the drug running planes to take out an armored car, or use a surface-to-air missile on a private jet. What you're talking about by "shooting them all" is starting another "war" where the "enemy" has the dramatic advantage of tons and tons of very liquid cash, worldwide connections, and a presence in every nation on Earth.
If your disk driver isn't formally verified, it can overwrite the boot sector.
If your video driver isn't formally verified, it can overwrite any location in memory.
If your (insert just about anything here) driver that supports DMA isn't formally verified, it can also overwrite any location in memory.
If your BIOS flashing driver isn't formally verified, the next time you boot you have a rootkit.
If your file system driver isn't formally verified, it can modify the operating system files.
If your window manager/login prompt/other common OS programs aren't formally verified, they will allow privilege escalation.
In short, everything the user has to trust when using a computer needs to be formally verified. That even includes web browsers and plugins that allow running untrusted code.
It doesn't matter how careful or professionally trained the application programmers are, nor how safe the programming language used to write the application is, when the OS isn't even designed to limit what they can do. All programs have bugs, you shouldn't have to trust them not to have them.
By extension, it doesn't matter how careful or professionally trained the operating system programmers are, because all operating systems have bugs. I completely agree.
It's time to start requiring formal verification of all system software so that it doesn't matter how bad the programmers are; if a simple proof verifier that is stored in the BIOS can't validate a proof of correctness at the machine code level for the boot sector it loads, it won't jump to it. The formally correct boot loader would then load the OS and its correctness proof and verify it before starting the OS. Part of the correctness proof would be showing that no modification of the BIOS, boot loader, or OS can occur unless the modification also has a valid proof that follows a security policy stored in the BIOS. Don't ever allow the initial proof verifier or the security policy to be modified. Prove the correctness (with regard to the axioms of set theory, at least) of the proof verifier's machine code in the same language, and the proof verifier can even verify itself every time it boots. Of course, you could claim that all mathematical proofs have bugs in them and that there's no way to trust the proof verifier to be correct. But then you might as well distrust ZF and most of mathematics by extension.
Re:Without remorse there is no rehabilitation.
on
Kevin Mitnick Answers
·
· Score: 1
Proof. I pay for movies. What I want is for movies to last longer than 5 minutes in a house with young kids who think DVDs make excellent coasters and frisbees. Thank the good ol' USA for making decss illegal.
Re:Without remorse there is no rehabilitation.
on
Kevin Mitnick Answers
·
· Score: 1
What I'm talking about is empathy. He's saying that he broke into computer systems, stole some information and terrorized them, but he didn't make a profit on it so it's ethically okay. That's bullshit. It's amoral. It's a complete lack of empathy, and a telling sign of a sociopath.
Why trust empathy when behavior is what counts? The best con men can generate as much false empathy as you want and still rob you blind. A sociopath would continue to pursue antisocial activities regardless of expressing empathy.
Re:Without remorse there is no rehabilitation.
on
Kevin Mitnick Answers
·
· Score: 1
I feel no remorse at being a criminal either. I feel remorse for living in a fascist society where making copies of digital information for personal use is a crime and you can go to jail for 2 years for having a smartphone in a theater.
If freedom is outlawed, only criminals will have freedom.
Yeah, but it's all binary trust. I fully trust Verisign, but I wouldn't even trust Digicert to sign hacker.ru. There's a problem with that, because I am only really about 60 to 70 percent sure that Verisign won't be compromised or sell out to China's interests, and I don't really care if hacker.ru is signed by a cheap root authority so long as I'm getting my cracks and wares from them and not evilhacker.ru. The current CA model is a forest of trust model. Every root CA forms a fully trusted tree, and every tree and branch and leaf in the forest is just as trusted as any other for all practical purposes. The forest needs fine-grained trust to become a web of trust, with cross-signing of individual certificates by many other certificate authorities. Then an overall trust measurement can be made.
The central support of ID is generally an argument that certain complex structures could not arise by naturalistic processes in the time-frames that modern evolutionary science has discovered. The arguments are generally broken down into statements about irreducible complexity (no viable evolutionary path between an original phenotype and a current phenotype) and specified complexity (the impossibility of creating information by random processes).
The former argument I attribute to a lack of imagination. For instance there are well-defined stages of eye evolution, and almost all intermediate stages are present in some species observable today. Many ID advocates nevertheless use the eye as an example of a structure that points to design, apparently ignoring the multitude of research on specific evolutionary models and examples from nature.
The latter argument is simply mathematically and physically incorrect. The basic definition of specified complexity almost exactly matches the criterion for compressibility; an example of a random letter, a random string of letters, and a string from a Shakespearean play is a good example. Guess which of the three is highly compressible and also used as an example of specified complexity. This should not be surprising because all living structures are essentially the result of "decompressing" the information stored in genes as they are repeatedly expressed or inhibited in complex feedback loops. It is trivial to create specified complexity by creating random formal grammars (context free or context sensitive, for example) and then generating random strings in the language by iteratively choosing a random production rule to extend a string of grammatical (terminal and non-terminal) symbols. It is trivial to generate random strings that look very much like Shakespeare (meter, word choice, etc.) but lack semantic meaning, yet I see no way of using the definition of specified complexity to distinguish between the two. A fitness function seems to be necessary to judge the semantic content, but given a fitness function over the language in which the strings are produced there exists an evolutionary processes to produce strings of ever higher fitness by randomly combining and changing the most fit strings that have been produced so far. Fit specified complex structures exist precisely because they can survive and reproduce, not because they are specified and complex.
Physically it can be shown that the sun is capable of converting its energy into tremendous amounts of information stored in the physical state of the Earth. Even if the Earth was a solid uniform sphere at some point in the past the repetitive heating and cooling by the sun would create more information (reduce local entropy) in the Earth over time by changing its configuration away from uniformity. The sun itself received its store of energy from an even more energetic star before it, and ultimately received its energy from the even more energetic big bang. Ultimately, ID must reduce to the sole argument that the laws of the universe and the initial conditions must have been intelligently designed. Given that even the rules of the game of life allow Turing completeness, it seems unlikely that the particular laws of our universe are necessary for sentient life. The anthropic principle suffices to explain why humans, and not aliens or glider-computers, are discussing this on slashdot. Given that the initial conditions of the big bang appear to be highly uniform, as well as the subsequent development of the universe, it seems likely that there is either no designer or that the designer is extremely subtle, to the point of being undetectable. Belief in a being because of an abundant lack of evidence does not, to me, seem scientific.
Slashdot Mirror
I said:
You can take a published string and make it a reasonably secure passphrase by adding enough entropy to it, but you still have to remember the entropy that you've added. Why not just start with a diceware passphrase and memorize the entropy directly?
I think that's still a valid point. How well can you misremember a quote? What is the maximum hamming distance between the original quote and a passphrase that you can remember? If you can remember 64 or more bits of entropy to add to an existing quote, you might as well remember a shorter diceware passphrase with the same entropy.
Additionally; how secret is your choice of source material? You can only have a finite number of books in your house, and a larger but finite number within driving distance. The likelihood of you traveling far and wide to generate a passphrase is pretty low. Can you be sure that Echelon didn't record the text (or at least the URLs) it's seen you fetch over your Internet connection? Reducing the search space to only a few thousand sources makes the problem almost embarrassingly simple. Build a probabilistic model of your writing/typing and then use it to find the nearest likely passphrases generated by altering the substrings of sources to better fit your writing style. Most likely you don't choose truly randomly from a set of altered quotes; you look for things in the text that seem easy to remember if they are changed, or that trigger some other memory that makes it easier to remember the other changes. Humans are bad at generating truly random text.
A 40 TB corpus has approximately 4 x 10^16 substrings of less than 1K, or 4 x 10^15 if we assume strings start on word boundaries and an average word length of 10 or less. Even if you charitably assume the whole hash/decrypt/validate process can be done in 1 ms of compute time per candidate, thats 4 x 10^13 seconds (about 1M years) of compute time. Unless Amazon has drastically lowered their prices, I don't think you'll be getting that for $150...
And in ten years, it will only be a thousand years of compute time using your numbers. Ten years later, only a year. Do your secrets only need to last 20 years? I would also remind you of GPU password cracking where billions of cryptographic primitives per second per card is not unusual. Assuming a thousand salting/setup primitive operations per passphrase (a bit lower than my preference for LUKS, but I've seen plenty of software with fewer iterations) would drop your estimate to 1000 GPU-years for an attacker to mount an attack today. Just pay some down-and-out bitcoin miners to put their now-overpriced rigs to use.
I admit that I underestimated the dollar amount for a well designed cryptosystem with iterated salting during key setup for dramatic effect. Taking some basic numbers from the bitcoin folks, it looks like a GPU cryptographic primitive (sha256 or ripemd160) costs around 1e-6 Joules. 4e16 substrings times 1000 primitives costs 4e13 Joules, or about 11 megawatthours, or about <pinky>1 million dollars</pinky>. However, there are probably many redundancies in the 40 TB database and it could be ordered by those redundancies to search through the more common space of text with shorter substrings first making it more likely to find weak passphrases quickly. For a weak cryptosystem where a single cryptographic operation suffices to test a passphrase I was only an order of magnitude off.
Fabrice Ballard already wrote an x86 emulator in javascript. Just install the standard x86 JVM inside of that and you're good to go.
What is the economic difference between an illegal alien invader and you?
Pays taxes? No difference.
Produces goods and services? No difference.
Consumes goods and services? No difference.
Possesses a particular piece of paper issued by the government at public expense? BINGO!
It's obviously foolish to use public text verbatim as a key. Common Crawl has a 40 TB dataset that costs approximately $150 to MapReduce on EC2. Any key that happens to be a (reasonably short, say under 1KB) substring of that data costs $150 to break. Any key within a short hamming distance of a substring in that database costs roughly 2^hamming_distance more to break; two changed bytes is only worth $600. I imagine that large organizations who care have much larger databases including the text of most published books. It's such an obvious idea and until you realize that attackers have access to all the public source data that you do it sounds like a good idea to just pick a random string from a book to use as a passphrase. Don't kid yourself; no matter how obscure or unpopular a song is there will be lyrics for it somewhere on the Internet, not to mention in published books.
You can take a published string and make it a reasonably secure passphrase by adding enough entropy to it, but you still have to remember the entropy that you've added. Why not just start with a diceware passphrase and memorize the entropy directly?
If you could translate any 3-SAT problem into a factoring problem in P time, then factoring would be NP-complete. Otherwise it's just in NP.
The problem with hyping a test of an emergency system is that its effectiveness drops nearly to zero around the time of the test. Even testing an alert system too often reduces its effectiveness. For a national alert system (and the September 11, 2001 attacks didn't even warrant an alert, apparently) it will probably end up being roughly useless. A propaganda tool at best.
A central authority doesn't need to be a fragile forest of fully trusted CAs like we have now. A much better solution would be for clients to have marginal trust in any individual x509 signature of a certificate, requiring at least N distinct signatures to validate the certificate, where N is great enough to significantly reduce the threat of enough compromised CAs signing an attacker's certificate to make it trusted. Inherited trust from an intermediate CA would only carry a portion of the trust placed in the root certificate of the chain so that each of the N signatures would have to come from independent organizations.
Basically, build a PGP-like web of trust out of the current CAs. It's even in the root CA's economic best interest because they'll sell N times as many certificates.
I think they are swamped with UNqualified candidates. Recruiting gives them a list of at least potentially qualified candidates. If their standard interview process has a false positive rate of 1% and they have thousands of candidates of whom 99% are unqualified, then 50% of their hires will be qualified. If their recruiters give them a list of candidates of whom 25% are qualified, their odds of a qualified hire go up to about 97%.
Juggling bits and juggling people are completely different skills.
The solution is just to turn people into bits with a chainsaw before juggling them.
People who have only ever used a single programming language and can't even understand another language are not programmers. I'm not sure they're even human. Same goes for application usage.
Just to be clear, I both believe in the existence of God (every possible god, in fact), and the non-existence of god at the same time. I exist in multiple universes, some with gods and some without, but enough of those universes are similar enough that I am unable to determine with sufficient confidence which universe this instance of myself is actually in.
Does this make me a theist, an atheist, or perhaps a schrodingerist?
In the interest of more accurate communication I'd like to introduce some other labels that should be applied to individuals as appropriate.
aspontaneous generationist (one who does not believe in spontaneous generation)
aplanar Earther (one who does not believe in a flat Earth)
acentralist (one who does not believe that the Earth is the center of the universe)
atroll (one who does not believe in trolling Internet discussion forums)
adead (one who believes they are still living)
aphlogistonist (one who does not believe in the phlogiston theory)
aplatonist (one who does not believe in the platonic solid arrangement of the Heavenly Spheres)
ageometer (one who does not believe that Euclid's Geometry is axiomatic)
asarcasmite (one who does not believe in using sarcasm in an argument)
Feel free to chime in with other vitally important labels that we can add to this list.
but every researcher who cites other works in their publications. This is a lawsuit against the process of science and scholarly review itself.
git's fast, cryptographically validated, and widely used and tested. Put everything you care about into a git repository, then clone it wherever you have extra space for backups. Enable remote pulls to keep the backup versions up to date.
I use subversion for the same purpose now, but I think git is overall a much better choice and I'll probably switch eventually.
The OS doesn't have to trust the boot sector, it can verify the information using cryptographic signatures. If you treat the block devices as untrusted, you will do things like using checksums, CRCs, ECC, etc... don't forget that hard drives typically get 1 out of 10^15 blocks wrong in the normal course of doing business.
DRM is not the same as formal correctness. I can probabilistically trust public key methods to protect probabilistically correct code, or I can fully trust formally proven code. Given the track record of DRM and code-signing I would opt for formal correctness. Hardware reliability is another issue entirely. You can only have probabilistic trust in the hardware, but unless you are working with tamper-resistant systems you probably don't have to worry about direct attacks on hardware reliability and can use ECC and cryptography to ensure that your own hardware is operating correctly.
Since an untrusted file system driver would have to use the OS to read / write the block devices, you could present it with a read-only capability to the block device which contains the OS, stopping all modification. In the case of a need for an update, the OS could then use a read/write capability to do the update, and switch back when done. I'm sure someone with a good background in CS could figure out a more secure way of doing it.
How does the OS know which data blocks contain user data and which blocks contain OS data? How does the OS know the filesystem isn't lying about the data it's going to write? Suppose the OS has perfect control over when and where the file system can write data. The file system says "Time to install an OS upgrade", gets permission, and then writes a rootkit instead of valid data. The entire chain has to be trusted. "a more secure way of doing it" is to formally verify the data to be written to the OS disks before it's written, and then when booting formally verify it again before running it to be sure it wasn't modified.
Not if the memory management is working correctly... Windows NT 3.5 and earlier had video running outside the kernel... which is the right way to do things in terms of security.
It's not just the CPU that has access to RAM, any device on the PCI bus has access as well. You do have to trust your hardware, of course, but once the hardware is trusted you also need to trust the driver software to give the correct addresses to the hardware for DMA. The hardware has no idea what sort of protection the OS is using or how it has organized memory, it just reads and writes to specific addresses in RAM. There were (are) numerous X11 privilege escalations based on this fact since OpenGL programs can talk directly to the hardware via direct rendering.
Drivers don't have to run in kernel mode to be efficient, and the DMA doesn't have to be set up directly by the drivers. The MMU and the control over input parameters to the DMA/Interrupt subsystems should suffice here as well.
Sometimes the DMA parameters are sent directly to the hardware, especially with video cards. DMA scatter/gather is also used by many disk and network controllers. The particular interface used to tell the hardware which addresses to use is not standardized to the point that every possible hardware driver could use a simple, secure operating system service to pass those DMA parameters to the hardware.
The primary reason we even worry about root kits is that Operating Systems haven't been designed to work in a world of untrusted code. Changing that one aspect of things, doing a hell of a lot of coding to build a capability based OS, provides an environment in which it is very unlikely that any rootkit would get the opportunity to be installed.
So you have a perfect capability operating system and it will never be infected by a rootkit and you have fine-grained control over network and file access. You have a secret file on your computer that you would like to process, the result of the proc
Maybe you underestimate the resources of these cartels. They are almost certainly capable of assassinating the Mexican president and any other members of government, and I would not be surprised if they could assassinate any political leader on Earth. It's not hard. Get your hands on an air-to-surface missile and use one of the drug running planes to take out an armored car, or use a surface-to-air missile on a private jet. What you're talking about by "shooting them all" is starting another "war" where the "enemy" has the dramatic advantage of tons and tons of very liquid cash, worldwide connections, and a presence in every nation on Earth.
If your disk driver isn't formally verified, it can overwrite the boot sector.
If your video driver isn't formally verified, it can overwrite any location in memory.
If your (insert just about anything here) driver that supports DMA isn't formally verified, it can also overwrite any location in memory.
If your BIOS flashing driver isn't formally verified, the next time you boot you have a rootkit.
If your file system driver isn't formally verified, it can modify the operating system files.
If your window manager/login prompt/other common OS programs aren't formally verified, they will allow privilege escalation.
In short, everything the user has to trust when using a computer needs to be formally verified. That even includes web browsers and plugins that allow running untrusted code.
It doesn't matter how careful or professionally trained the application programmers are, nor how safe the programming language used to write the application is, when the OS isn't even designed to limit what they can do. All programs have bugs, you shouldn't have to trust them not to have them.
By extension, it doesn't matter how careful or professionally trained the operating system programmers are, because all operating systems have bugs. I completely agree.
It's time to start requiring formal verification of all system software so that it doesn't matter how bad the programmers are; if a simple proof verifier that is stored in the BIOS can't validate a proof of correctness at the machine code level for the boot sector it loads, it won't jump to it. The formally correct boot loader would then load the OS and its correctness proof and verify it before starting the OS. Part of the correctness proof would be showing that no modification of the BIOS, boot loader, or OS can occur unless the modification also has a valid proof that follows a security policy stored in the BIOS. Don't ever allow the initial proof verifier or the security policy to be modified. Prove the correctness (with regard to the axioms of set theory, at least) of the proof verifier's machine code in the same language, and the proof verifier can even verify itself every time it boots. Of course, you could claim that all mathematical proofs have bugs in them and that there's no way to trust the proof verifier to be correct. But then you might as well distrust ZF and most of mathematics by extension.
Oh, wait.
Proof. I pay for movies. What I want is for movies to last longer than 5 minutes in a house with young kids who think DVDs make excellent coasters and frisbees. Thank the good ol' USA for making decss illegal.
What I'm talking about is empathy. He's saying that he broke into computer systems, stole some information and terrorized them, but he didn't make a profit on it so it's ethically okay. That's bullshit. It's amoral. It's a complete lack of empathy, and a telling sign of a sociopath.
Why trust empathy when behavior is what counts? The best con men can generate as much false empathy as you want and still rob you blind. A sociopath would continue to pursue antisocial activities regardless of expressing empathy.
I feel no remorse at being a criminal either. I feel remorse for living in a fascist society where making copies of digital information for personal use is a crime and you can go to jail for 2 years for having a smartphone in a theater.
If freedom is outlawed, only criminals will have freedom.
Yeah, but it's all binary trust. I fully trust Verisign, but I wouldn't even trust Digicert to sign hacker.ru. There's a problem with that, because I am only really about 60 to 70 percent sure that Verisign won't be compromised or sell out to China's interests, and I don't really care if hacker.ru is signed by a cheap root authority so long as I'm getting my cracks and wares from them and not evilhacker.ru. The current CA model is a forest of trust model. Every root CA forms a fully trusted tree, and every tree and branch and leaf in the forest is just as trusted as any other for all practical purposes. The forest needs fine-grained trust to become a web of trust, with cross-signing of individual certificates by many other certificate authorities. Then an overall trust measurement can be made.
The central support of ID is generally an argument that certain complex structures could not arise by naturalistic processes in the time-frames that modern evolutionary science has discovered. The arguments are generally broken down into statements about irreducible complexity (no viable evolutionary path between an original phenotype and a current phenotype) and specified complexity (the impossibility of creating information by random processes).
The former argument I attribute to a lack of imagination. For instance there are well-defined stages of eye evolution, and almost all intermediate stages are present in some species observable today. Many ID advocates nevertheless use the eye as an example of a structure that points to design, apparently ignoring the multitude of research on specific evolutionary models and examples from nature.
The latter argument is simply mathematically and physically incorrect. The basic definition of specified complexity almost exactly matches the criterion for compressibility; an example of a random letter, a random string of letters, and a string from a Shakespearean play is a good example. Guess which of the three is highly compressible and also used as an example of specified complexity. This should not be surprising because all living structures are essentially the result of "decompressing" the information stored in genes as they are repeatedly expressed or inhibited in complex feedback loops. It is trivial to create specified complexity by creating random formal grammars (context free or context sensitive, for example) and then generating random strings in the language by iteratively choosing a random production rule to extend a string of grammatical (terminal and non-terminal) symbols. It is trivial to generate random strings that look very much like Shakespeare (meter, word choice, etc.) but lack semantic meaning, yet I see no way of using the definition of specified complexity to distinguish between the two. A fitness function seems to be necessary to judge the semantic content, but given a fitness function over the language in which the strings are produced there exists an evolutionary processes to produce strings of ever higher fitness by randomly combining and changing the most fit strings that have been produced so far. Fit specified complex structures exist precisely because they can survive and reproduce, not because they are specified and complex.
Physically it can be shown that the sun is capable of converting its energy into tremendous amounts of information stored in the physical state of the Earth. Even if the Earth was a solid uniform sphere at some point in the past the repetitive heating and cooling by the sun would create more information (reduce local entropy) in the Earth over time by changing its configuration away from uniformity. The sun itself received its store of energy from an even more energetic star before it, and ultimately received its energy from the even more energetic big bang. Ultimately, ID must reduce to the sole argument that the laws of the universe and the initial conditions must have been intelligently designed. Given that even the rules of the game of life allow Turing completeness, it seems unlikely that the particular laws of our universe are necessary for sentient life. The anthropic principle suffices to explain why humans, and not aliens or glider-computers, are discussing this on slashdot. Given that the initial conditions of the big bang appear to be highly uniform, as well as the subsequent development of the universe, it seems likely that there is either no designer or that the designer is extremely subtle, to the point of being undetectable. Belief in a being because of an abundant lack of evidence does not, to me, seem scientific.