This is the Government Security Program, through which they release the source code of Windows versions to governments around the world, obviously including USG, but also including Russia. Windows 10 isn't on this list at the moment, but 8.1 and 7 are, and one is pretty safe in assuming that nothing of note has changed here, and DoD will have full source code access JUST AS THEY ALWAYS HAVE.
Further, they often DO get customizations to their deliverables, so it's almost ASSURED that their version won't be the same that the rest of us have access to. In FACT the article even explicitly mentions that its ENTERPRISE in the first place- you know, the version you can't buy, because you aren't a corporation and therefore have no expectation of privacy. Even enterprise seems to still have issues with chatting, but it's the ONLY version (unlike, say, Pro) where you can in THEORY set telemetry to "none".
So BOTH points are correct- the government won't be using the same version as "us plebes" (for two reasons) and ALSO they will have source code.
And for making me log in, lemme second the "go shill on Ars" guy. Good grief.
No. This is not one of your best bets or security, because it doesn't mean this TO YOU. The premise of my post is what you, iphone owner, can do now that we find that it's possible for Apple to be coerced into writing (or for the Chinese government to write and sign with Apple's key) a malicious ios. You don't have the ability to implement this excellent two-stage security solution, so it's not something useful to you. You can choose a quick to type 8-ish character passphrase (not using words, of course) and get in and out of your phone about as easy as with a PIN, and then you are safe as long as the hardware doesn't fall, or you can choose a seriously real passphrase like you would for full disk encryption, and then you are safe as long as their implementation isn't shit.
A lot of applications demand a hosts file already, because a DNS can be compromised more easily than an IP can be spoofed. Obviously this isn't a general solution for a generic server or useful desktop.
> Breaking *anything* is a matter of cost and willingness.
AES-256? Serpent 256? Twofish 256?
The goal of crypto is to make something unbreakable. It seems it has generally succeeded, as best we know, as it appears impossible to do it. Brute forcing a 256 bit key is not a matter of cost. Getting into some hardware probably is, coercing software (if the ability to push the install remotely exists) absolutely is.
The order implies that Apple is capable of delivering a remote update, or that forcing an update locally is possible if you have physical access. It also implies that portions of the security models are enforced by software that is vulnerable to "update", such as the wipe-after-ten-tries (presumably that code will be replaced with a no-op) and the code entry delay in excess of that which is enforced by hardware.
Whether Apple is compelled to do this or not, the natural concern is "well how much of my data is shielded by math, how much by hardware, and how much by software"?
You can't bargain with math, you have a devil of a time working out hardware, and software along is meaningless as a defense.
It appears that your best bet for security is either:
1)- A multi-character password that is easy to enter (and you'll remember it if its your phone password, lol), but reasonably short. This is if you trust that the 80ms hardware delay can't be broken. This precludes the use of 4 and 6 digit PINs, as a 4 digit PIN will usually fall after a few minutes of this treatment, and a 6 digit PIN after around half a day. An 8 digit password consisting of a completely random set of just the visible lowercase letters (aka, no actual english words) at this rate is hundreds of years, and adding stuff that's harder to enter quickly (capitals, numbers, special characters) makes it much more secure, as does lengthening the password slightly. The challenge here is that passwords are usually chosen to be words, greatly reducing the entropy. And again, this assumes that the 80ms hardware delay is not defeatable.
2)- A fully secure crypto passhprase. This is the level of drama you would go through to password protect a drive or something you take very seriously, and as such it would be a lot more than 8 characters. Your passphrase is long, contains several unpredictable parts, and makes use of more than just a statistically predictable subset of words and characters. You can set this on the iphone, of course, but this kind of protection is not trivial to type in. In this case, you are trusting the math only, however, and assuming that the software will be compelled by the government, and the hardware will be owned by a team skilled in this matter.
Going forward, Apple should probably move the "erase after 10 tries" into the secure portion of the phone, such that it has a protected portion that can't be overwritten without access to the PIN. This will also make them immune to this sort of order in the future.
The Apple post seems to make it clear that this attack is doable. This implies that the San Bernadino phone is set to gobble up an update signed by Apple or something. This sounds like Apple is fully capable of creating a signed malicious operating system into the phone, and is now refusing to do so. Nothing in the article implies that this is *impossible*, merely *really unwise*.
Of course it is unwise. But the fact that it is POSSIBLE belies a second security flaw- that installing a new OS is possible to anyone without the PIN. That's a security flaw, and it means that breaking any iphone is now a matter of cost and willingness, not possibility.
I could be reading this wrong, and its not directly stated, but that's the implication, at least?
I've put this elsewhere in the thread but Apple seems to think they can provide plenty of stuff from icloud to law enforcement. The icloud stuff is encrypted with a passcode known to Apple:
So if it was in icloud, presumably they have it already, because Apple says "we can give you the icloud stuff, because we can access it". The locally encrypted stuff is locally encrypted, however- so presumably they want access to that.
> While Slashdot has traditionally relied on user moderation, it would be hard to find many people in defense of APK.
I feel that he is fighting the good fight by having a piece of software that is well intentioned and appears to be functional. The fact that he would fill up a comment page back in the day with the same copy-pasta is indefensible, of course. I've argued with him a few times (there are weaknesses to host based blocking), and I never felt that he was a bad guy. He toned down his paste-rate in the last few months as well, but it was still disruptive and clearly required moderators to go through and delete manually. It seems whipslash has a more permanent fix.
The other post explained the error, but the reason for this is because the entropy in the key would be very low or guessable, so the break-in procedure would be (1) image the phone (2) try every PIN. You'd be in in less than a second. The key is a 128 bit AES key (some posts claim 256, so maybe that's correct, but I thought it was 128), and that's the piece that is guarded by the PIN. All the shenanigans about auto-wiping and machine enforced attempt limits are to allow the use of such a low entropy password like a PIN in the first place, by being able to wipe the master key if the user can't input the correct code in ten tries.
The first is a 4 digit PIN. The second is a 6 digit PIN. The third is any passphrase of any length.
It's trivially obvious which mode it is in- the first two bring up a number pad and have 4 or 6 boxes to fill in, the third brings up a screen with a keyboard.
> Were I so desperate to get into the phone, I'd image it
Right, so now you have an AES-128 image sitting around, and you destroyed the key when you imaged it. Unless they dicked up the AES-128, it should be pretty hard to break that. The key in question isn't the PIN, obviously, the PIN protects the key.
There's a part where the document sort of complains that users aren't required to back everything up to icloud, because they can just ask for anything in icloud at all and get it in plaintext immediately (as documented by the first link).
If you promise to encrypt "hunter2" your end with AES-256, is it encrypted? Sure, but it's also here on plaintext, in transit, and if asked, you could certainly retrieve it. Even though it's clearly my password that you can't see:P
It's possible he turned Siri off on lockscreen. It's much more possible that the few things accessible in that manner aren't what they are looking for.
You can turn the feature off- but under what circumstances would you want someone to have access to your phone for two hours and it continue to have all your personal stuff on it?
Remember that restoring an iphone is trivial once its in your possession, from itunes or icloud.
"We provide a product that works as advertised, and it can't be broken into" might be slammed by some pundits, but it's certainly not going to make them look bad to their potential customers.
There aren't enough terrorists to make a difference in sales.
However, I will say this: if Apple can break in, it makes me less likely to trust anything Apple says or sells, because I expect encryption, not backdoored bullshit.
But it has to result in a DNS inquiry (so if your addresses are stored locally, you're ok), and it has to go to a malicious DNS server or somehow be served a malicious DNS packet in response.
The proprietary drivers (and firmware) include agreements that nvidia made with other companies, so they can't open source them. The problem here isn't even the existence of the firmware that was always loaded, it was that the normal workaround no longer worked, because nvidia was seeing scammers loading firmware for more advanced cards on less advanced cards- this didn't improve the cards, but it DID let them lie about what they were. I'm glad they are coming up with a solution for the open source stuff. But like anyone trying to game in Linux, I'm using the proprietary driver ball anyway:/
> We have been promised AI was just around the corner since the 50s and the 60s.
Like fusion, AI has been twenty years away for like two generations. Doesn't matter though.
Now, we may actually be close this time- but that's not important. What's important for this conversation isn't strong AI. You don't need to be able to take the job of a poet or president to absolutely disrupt economies. You don't need strong AI. You just need AI, and frankly, you often don't even need that- just good programs and fast hardware.
> Show me even one system that is able to match human beings in creativity and resourcefulness.
Do you think that most jobs require creativity and resourcefulness? Your example is the LIGO experiment. Do you think the men there are representative of typical human jobs? Ok, so top research scientists aren't gonna be replaced by AI any time soon. But even if you are one of these creative top tier people, you have to recognize that most jobs aren't. Once a truck is certified to be driven by AI with a call center that can override in weather, that will blast away like half of trucking jobs within five years. Once people are in ANY way used to placing an order automatically at McDonalds, that will blow up huge numbers of jobs across the service industry.
You don't need an AI capable of dreaming and launching Von Neumann probes across the local group for this- you need what we have now with either a little or a lot of raw software development thrown at it.
>But HAL is definitely not in our future. Are you sure? I wouldn't place a bet saying that, but I dunno if I'd place the opposite bet either.
>And no sex robots either. There's already sex robots. And as you might imagine, the AI isn't really the limiting tech on them, and CERTAINLY not the overall limiter on adoption. If you wanted to make and sell sex robots, you have some AI tasks to solve, but mostly you need to solve a combination of robotics and materials, and then you need to somehow convince everyone that having some twenty thousand dollar fuck doll is not at all creepy and fucked up, in addition to selling them on it being a good idea. The sex doll angle is just there to make you click.
That's the standard narrative. Another standard one counters with "well everyone will just get jobs maintaining and building robots". But there's a bunch of OTHER ways it could go- you could use this to justify a great deal of different political moves, from minor to extreme, in more than one direction.
Slashdot Mirror
INCORRECT! And you made me log in.
http://download.microsoft.com/...
This is the Government Security Program, through which they release the source code of Windows versions to governments around the world, obviously including USG, but also including Russia. Windows 10 isn't on this list at the moment, but 8.1 and 7 are, and one is pretty safe in assuming that nothing of note has changed here, and DoD will have full source code access JUST AS THEY ALWAYS HAVE.
Further, they often DO get customizations to their deliverables, so it's almost ASSURED that their version won't be the same that the rest of us have access to. In FACT the article even explicitly mentions that its ENTERPRISE in the first place- you know, the version you can't buy, because you aren't a corporation and therefore have no expectation of privacy. Even enterprise seems to still have issues with chatting, but it's the ONLY version (unlike, say, Pro) where you can in THEORY set telemetry to "none".
So BOTH points are correct- the government won't be using the same version as "us plebes" (for two reasons) and ALSO they will have source code.
And for making me log in, lemme second the "go shill on Ars" guy. Good grief.
No. This is not one of your best bets or security, because it doesn't mean this TO YOU. The premise of my post is what you, iphone owner, can do now that we find that it's possible for Apple to be coerced into writing (or for the Chinese government to write and sign with Apple's key) a malicious ios. You don't have the ability to implement this excellent two-stage security solution, so it's not something useful to you. You can choose a quick to type 8-ish character passphrase (not using words, of course) and get in and out of your phone about as easy as with a PIN, and then you are safe as long as the hardware doesn't fall, or you can choose a seriously real passphrase like you would for full disk encryption, and then you are safe as long as their implementation isn't shit.
That would be a wicked cool design though.
A lot of applications demand a hosts file already, because a DNS can be compromised more easily than an IP can be spoofed. Obviously this isn't a general solution for a generic server or useful desktop.
> Breaking *anything* is a matter of cost and willingness.
AES-256? Serpent 256? Twofish 256?
The goal of crypto is to make something unbreakable. It seems it has generally succeeded, as best we know, as it appears impossible to do it. Brute forcing a 256 bit key is not a matter of cost. Getting into some hardware probably is, coercing software (if the ability to push the install remotely exists) absolutely is.
The order implies that Apple is capable of delivering a remote update, or that forcing an update locally is possible if you have physical access. It also implies that portions of the security models are enforced by software that is vulnerable to "update", such as the wipe-after-ten-tries (presumably that code will be replaced with a no-op) and the code entry delay in excess of that which is enforced by hardware.
Whether Apple is compelled to do this or not, the natural concern is "well how much of my data is shielded by math, how much by hardware, and how much by software"?
You can't bargain with math, you have a devil of a time working out hardware, and software along is meaningless as a defense.
It appears that your best bet for security is either:
1)- A multi-character password that is easy to enter (and you'll remember it if its your phone password, lol), but reasonably short. This is if you trust that the 80ms hardware delay can't be broken. This precludes the use of 4 and 6 digit PINs, as a 4 digit PIN will usually fall after a few minutes of this treatment, and a 6 digit PIN after around half a day. An 8 digit password consisting of a completely random set of just the visible lowercase letters (aka, no actual english words) at this rate is hundreds of years, and adding stuff that's harder to enter quickly (capitals, numbers, special characters) makes it much more secure, as does lengthening the password slightly. The challenge here is that passwords are usually chosen to be words, greatly reducing the entropy. And again, this assumes that the 80ms hardware delay is not defeatable.
2)- A fully secure crypto passhprase. This is the level of drama you would go through to password protect a drive or something you take very seriously, and as such it would be a lot more than 8 characters. Your passphrase is long, contains several unpredictable parts, and makes use of more than just a statistically predictable subset of words and characters. You can set this on the iphone, of course, but this kind of protection is not trivial to type in. In this case, you are trusting the math only, however, and assuming that the software will be compelled by the government, and the hardware will be owned by a team skilled in this matter.
Going forward, Apple should probably move the "erase after 10 tries" into the secure portion of the phone, such that it has a protected portion that can't be overwritten without access to the PIN. This will also make them immune to this sort of order in the future.
The Apple post seems to make it clear that this attack is doable. This implies that the San Bernadino phone is set to gobble up an update signed by Apple or something. This sounds like Apple is fully capable of creating a signed malicious operating system into the phone, and is now refusing to do so. Nothing in the article implies that this is *impossible*, merely *really unwise*.
Of course it is unwise. But the fact that it is POSSIBLE belies a second security flaw- that installing a new OS is possible to anyone without the PIN. That's a security flaw, and it means that breaking any iphone is now a matter of cost and willingness, not possibility.
I could be reading this wrong, and its not directly stated, but that's the implication, at least?
I've put this elsewhere in the thread but Apple seems to think they can provide plenty of stuff from icloud to law enforcement. The icloud stuff is encrypted with a passcode known to Apple:
http://www.apple.com/privacy/d...
This LEO guide seems to back that up:
http://manhattanda.org/sites/d...
So if it was in icloud, presumably they have it already, because Apple says "we can give you the icloud stuff, because we can access it". The locally encrypted stuff is locally encrypted, however- so presumably they want access to that.
> While Slashdot has traditionally relied on user moderation, it would be hard to find many people in defense of APK.
I feel that he is fighting the good fight by having a piece of software that is well intentioned and appears to be functional. The fact that he would fill up a comment page back in the day with the same copy-pasta is indefensible, of course. I've argued with him a few times (there are weaknesses to host based blocking), and I never felt that he was a bad guy. He toned down his paste-rate in the last few months as well, but it was still disruptive and clearly required moderators to go through and delete manually. It seems whipslash has a more permanent fix.
"Can your new site do 16 things?"
The other post explained the error, but the reason for this is because the entropy in the key would be very low or guessable, so the break-in procedure would be (1) image the phone (2) try every PIN. You'd be in in less than a second. The key is a 128 bit AES key (some posts claim 256, so maybe that's correct, but I thought it was 128), and that's the piece that is guarded by the PIN. All the shenanigans about auto-wiping and machine enforced attempt limits are to allow the use of such a low entropy password like a PIN in the first place, by being able to wipe the master key if the user can't input the correct code in ten tries.
Isn't this the exact attack that physical anti-tamper is meant to defeat?
You have three options:
The first is a 4 digit PIN.
The second is a 6 digit PIN.
The third is any passphrase of any length.
It's trivially obvious which mode it is in- the first two bring up a number pad and have 4 or 6 boxes to fill in, the third brings up a screen with a keyboard.
> Were I so desperate to get into the phone, I'd image it
Right, so now you have an AES-128 image sitting around, and you destroyed the key when you imaged it. Unless they dicked up the AES-128, it should be pretty hard to break that. The key in question isn't the PIN, obviously, the PIN protects the key.
NO!
If he had it on icloud, Apple could turn it over. The icloud backups are encrypted BY APPLE.
Check page 4:
www.apple.com/privacy/docs/legal-process-guidelines-us.pdf
Here's some guidelines:
http://manhattanda.org/sites/d...
There's a part where the document sort of complains that users aren't required to back everything up to icloud, because they can just ask for anything in icloud at all and get it in plaintext immediately (as documented by the first link).
If you promise to encrypt "hunter2" your end with AES-256, is it encrypted? Sure, but it's also here on plaintext, in transit, and if asked, you could certainly retrieve it. Even though it's clearly my password that you can't see :P
It's possible he turned Siri off on lockscreen. It's much more possible that the few things accessible in that manner aren't what they are looking for.
Actually, it will only take 2 hours before it has had enough tries to fire the 10-tries-delete-AES-key failsafe. Then it's gone for good.
You can turn the feature off- but under what circumstances would you want someone to have access to your phone for two hours and it continue to have all your personal stuff on it?
Remember that restoring an iphone is trivial once its in your possession, from itunes or icloud.
Anyway, it's not on by default.
> Mine isn't set to wipe after 10.
Easily fixable. Settings -> Touch ID and Passcode -> Turn on "Erase Data" at the bottom.
"We provide a product that works as advertised, and it can't be broken into" might be slammed by some pundits, but it's certainly not going to make them look bad to their potential customers.
There aren't enough terrorists to make a difference in sales.
However, I will say this: if Apple can break in, it makes me less likely to trust anything Apple says or sells, because I expect encryption, not backdoored bullshit.
The idea that a judge doesn't understand technology is NOT absurd, however.
> hybrid limited/unlimited
The word for that is "limited". If I can find a limit, it's "limited". It's not "hybrid limited/unlimited". That's silly.
But it has to result in a DNS inquiry (so if your addresses are stored locally, you're ok), and it has to go to a malicious DNS server or somehow be served a malicious DNS packet in response.
The proprietary drivers (and firmware) include agreements that nvidia made with other companies, so they can't open source them. The problem here isn't even the existence of the firmware that was always loaded, it was that the normal workaround no longer worked, because nvidia was seeing scammers loading firmware for more advanced cards on less advanced cards- this didn't improve the cards, but it DID let them lie about what they were. I'm glad they are coming up with a solution for the open source stuff. But like anyone trying to game in Linux, I'm using the proprietary driver ball anyway :/
> We have been promised AI was just around the corner since the 50s and the 60s.
Like fusion, AI has been twenty years away for like two generations. Doesn't matter though.
Now, we may actually be close this time- but that's not important. What's important for this conversation isn't strong AI. You don't need to be able to take the job of a poet or president to absolutely disrupt economies. You don't need strong AI. You just need AI, and frankly, you often don't even need that- just good programs and fast hardware.
> Show me even one system that is able to match human beings in creativity and resourcefulness.
Do you think that most jobs require creativity and resourcefulness? Your example is the LIGO experiment. Do you think the men there are representative of typical human jobs? Ok, so top research scientists aren't gonna be replaced by AI any time soon. But even if you are one of these creative top tier people, you have to recognize that most jobs aren't. Once a truck is certified to be driven by AI with a call center that can override in weather, that will blast away like half of trucking jobs within five years. Once people are in ANY way used to placing an order automatically at McDonalds, that will blow up huge numbers of jobs across the service industry.
You don't need an AI capable of dreaming and launching Von Neumann probes across the local group for this- you need what we have now with either a little or a lot of raw software development thrown at it.
>But HAL is definitely not in our future.
Are you sure? I wouldn't place a bet saying that, but I dunno if I'd place the opposite bet either.
>And no sex robots either.
There's already sex robots. And as you might imagine, the AI isn't really the limiting tech on them, and CERTAINLY not the overall limiter on adoption. If you wanted to make and sell sex robots, you have some AI tasks to solve, but mostly you need to solve a combination of robotics and materials, and then you need to somehow convince everyone that having some twenty thousand dollar fuck doll is not at all creepy and fucked up, in addition to selling them on it being a good idea. The sex doll angle is just there to make you click.
That's the standard narrative. Another standard one counters with "well everyone will just get jobs maintaining and building robots". But there's a bunch of OTHER ways it could go- you could use this to justify a great deal of different political moves, from minor to extreme, in more than one direction.