Slashdot Mirror
Intel to Build Encryption Capabilities in Chips
Will Johnston sent us
a link to a CNNfn article where you can read about Intel's
plan's to incorporate encryption into
it's microchips. I'm not sure about this one: The paranoid
in me starts quivering, but then again, standard encryption
sure would make a lot of this stuff a lot easier. What
do you think?
Careful how you read the article. It really doesn't tell you anything about what Intel is planning. Other articles I've seen through Slashdot would give you a better clue to what is going on...
Depending on how it's done, fast decryption inside the processor would allow the processor to execute an encrypted instruction stream directly, making it virtually impossible for anybody to figure out what the code does or to run it on another processor (or even an upgraded machine). It would then also make it much easier for companies to distribute encrypted digital content that is very difficult to copy or use even within the fair use provisions of copyright law.
... to have some accelerated math function like the one used in the RSA algorithm.
I hope they don't plan a high level funtion like RSAEncrypt(PrivatKey), this would fall under US encryption laws, opening the door to US Chips smuggling, new low-level software tweeks (this could be fun actually) and so on.
CU Guys
...well, if you know people at intel that is! ;-)
;-) hehehe..
a friend of mine who works at intel's architecture labs told me that each next-generation-CPUs (merced mostlikely) will containt a "burned-in" unique ID# and hardware random number generator (RNG).
what's gona happen if all CPU makers jump on the bandwagon? more secure e-commerce and non-existant pirating. (cuz we'll be selling software licenses by CPU ID#s!)
i think they're shooting themselves in the foot.. if you tie the value of a CPU to the software that it's lisenced to run, then you'll never upgrade an XT running a 3 MILLION $ software!
GOD i hope AMD dosen't do the same.. imagine a few years from now, when (as Larry Elison thinks) we'll all be buying our software online, after you pay, the site asks you "would you like to download your 'locked' intel version or your 'honor-system' amd version?"
Somebody please tell me this isn't happening.
I remember reading in sci.crypt in July (I think) that Intel was ordered to put backdoors in their chips by the NSA or something (if anyone has the post please put it up here). At the time I just blew it off as more paranoid delusions, but now...
Of course there was also someone else that posted that the NSA had a patent on a child's car seat (this was posted around the same time as the other post; again if anyone has it or finds it on DejaNews, I'd like to see it).
Jeph
themadfish[at]usa.net
Actually, Andy Grove leaked the P# info. But yes, AC is nice. Just not when you know your transactions are logged by the company network.
Software encryption is fast enough for most uses (and much more flexible). I don't see the point in doing it in hardware, unless the computer will be devoted to doing encryption only.
the primary motivation is to make software piracy almost impossible. Software is delivered encrypted for your particular chip ID, and is only decrypted as it is executed, instruction by instruction, inside the CPU. My source for this is the cypherpunk Tim May, who made his fortune at Intel. The key words in the press release were Built-in encryption could also make it easier to add access controls. As another coward observed above, this is old news.
Every(?) piece of hardware has a serial number that can be tracked over the internet. It's been used to find stolen componants like memory chips. In any case it's just a glorified dongle if they try to use it for anti-piracy. Just another protection to be cracked by people smarter than me. Anonymous Coward? You betcha. With a big capitol "C".
How long does your computer sit there idling? how much time does it take for an encryption exchange?
I can garantee that this idea is not for the 'consumer' but in the interest of intel and/or the us goverment. If speed was an issue we would simply have a separate encoder/decoder chip. e-commerce blah, don't be fooled.
-kojak
1 : iNTEL is a US company.
2 : US companys can onle export week encription.
3 : iNTEL won't eaven begin tpo add the overhead of disiding in advance which chips go where.
4 : There is no need for encription in hardware. software cripto is fast enogh eaven at ridiculus levles of security and insane key sizes ( 256 bit DES anyone ? )
ENCRIPTED software can't be moved from one PC to the next.
Conclusion. this was designed so that when you upgrade your machine you have to buy all new software. Microsoft wanteds this. MS got it and the DOJ had better pay attention to this particular case since iNTEL gains nothing.
Lots of people think that just because software
copy protection has failed in the past, it will
fail in the future. Maybe. But so far the
missing link has been universal hardware support
for authentication. If that becomes a reality,
it will be a different world.
The record companies will love this. So will
publishers who want pay-per-view.
The encryption functions that they build in may be bunk, but if it includes an accessable hardware random number generator which produces good entropy, I might just buy one. This from someone who has not *ever* owned an intel processor, and hasn't had a clone of one since the early days of the 486. (sparcs kick ass.)
Generating good random numbers is basicly impossible in software, and although I wouldn't trust the intel stuff too much, it would probably be good for one bit of entropy per ten bits of output, or something like that, as input to a software system that I trust.
Forget about common and cheap hardware that gives you "good entropy". Random generators is the most vulnerable part in best public key cryptosystems,
if NoSuchAgency pushes so hard for low bits crypto, do you think they will let you have a GOOD SOURCE OF ENTROPY??? Forget it. No chance.
Even if they will claim their hardware uses white noise as source for entropy how the heck you will be able to check it???? We don't even have a clue about what really CPU does (person who has
listings of the microcode please speak up). Hardware is not really an open source. Only folks from reverse engeneering teams at AMD and such will know what really is in these CPUs, oops, did I say they reverse engeneer which is against license??? bummer, yet another military secret out, but well, they ALL do it anywayz.
Yes, mixing the stuff provided by the intel chip with the standard /dev/random stuff is what I was thinking of.
/dev/random scheme is about the best I have seen, however it doesn't look like (and no, I am not sure) it would hold up so well on a mostly idle system. Some dedicated hardware support would give it a real boost.
And yes, the linux
Speaking of /dev/random, does anyone know if the /dev/random implementation under solaris 2.[567] is any good? Does it use mostly the same methods as the linux version?
Too slow, Intel! IBM already beat you to the punch. Their implementation speeds up secure electronic transactions on S/390 servers...
I notice many people are clamoring about the hardware vs software encryption issue. Some companies are implementing their devices in programmable logic... allowing you to "reconfigure" the device at any time. Perhaps this is more appropriate for encryption hardware?
what's gona happen if all CPU makers jump on the bandwagon? more secure e-commerce and non-existant pirating. (cuz we'll be selling software licenses by CPU ID#s!)
:-)
Yep, then everyone's gonna use this free "open source software" - what a horrible idea...
There are a few operations that are much faster in silicon than in software, like modular exponentiation. See the nFast hardware for example of a good way to accelerate crypto.
Most Hardware systems that need to generate random numbers use something called an LFSR (linear Feedback Shift Registers). These are trivial to implement in hardware (or reconfigurable hardware eg. FPGA, CPLD, etc), but are a pain in the ass to compute in software. There are plenty of books written on LFSR's and their properties. Bottom Line is that Intels is probably just planning on adding some long (>42 bits) LFSR on their chips will programmable masks (a mask offsets the bit sequence)
btw, what was so wrong with the clipper chip anyway!? the government dosen't care about me, i'm a nobody!! i'm more scared of my next door neighbor listening to my cordless phone conversations!! :-(
All it would take is one more terrorist bombing and encryption is GONE. That will be all the ammunition the government needs to pass stricter legislation for encryption and weapons.
And what congressman would vote against this new legislation created to "prevent further incedents from occuring" - such a congressman would be labeled a terrorist! THAT'S HOW WE LOSE OUR RIGHTS PEOPLE!!!! And the American public plays a big role in this with their ignorance!
I dont mean to insult anyone's intelligence here, as I'm sure most of you already know this.
ouch, can you imagine a cookie that would store your CPU-ID#? grrrrrrrr............ *bigbrother!*
No thanks. Gimme gnupg any day...
--
--
Just lurking, thanks!
Can be found at the Intel press release site. It appears that Intel will be working with RSA on incorporating RSA technology such as BSafe.
This actually makes me feel marginally better. RSA is sometimes fairly good about publicizing their algorithms. They are a pretty reasonable player in the encryption marketplace.
...phil
"For a list of the ways which technology has failed to improve our quality of life, press 3."
Encryption should be implemented as a software module of some sort. That way, if the algorithm gets broken, the module can be swapped out and replaced with a newer module that hasn't been cracked yet. I believe this is already done.
Posted by Bill, the Galactic Hero:
;) oooooooaaaaaaaaaayyyy
Yuggek muh duhdordik, cnuup buhyoyay hmerp. Gwee, fubbub spinkmadoodink -- qwrrr xoxop. Splunggi jojingloo meezoom:
?) thukthukthuk
#) krebbler gnay "inihimijimipibi"
- Bonk
no matter what encryption they use, it'll have to be considered weak. there isn't a SINGLE case in history where "security through obscurity" actually worked.
Assorted stuff I do sometimes: Lemuria.org
... if the algorithm they use isn't a standard one, you'll be forced to use Intel chips to use the algorithm. This says nothing, of course, about "is the algorithm 'secure'?" ...
also, if after a few years, the algorithm becomes cheap to brute force, you'll be required to upgrade.
so, in summary: encryption in the chip? feh! chips are fast enough today to handle software encryption. it tends to be more flexible as well.
I heard on headline news that the government will be able to see what was encrypted.
No thanks.
----
First midi was exported from software to hardware, then 3D graphics were exported to hardware, then device dependant drivers were exported from the OS to the hardware. Now encryption is exported from software to the hardware. What does it mean for the future of the programmer?
Easier is N E V E R more secure.
I still use PGP v2.06 because I know 5.0 has the govt. backdoor built in that Zimmerman had to give to the feds if he was going to be allowed to sell it...although this GNUgp sounds promising..
It's sure to be government-breakable, ie insecure.
If hardware encryption becomes the standard this will kill off any Open Source encryption momentum. You'll note that the Wassenaar agreement specifically doesn't address Open Source. see:
www.gnupg.org. So if everyone gets hooked on hardware encryption it's one step closer on the slippery slope to 'clipper' type key escroe our friends at the NSA have been pushing all decade.
I won't be buying into that garbage, nor would I trust Intel or any other huge corporation with my privacy.
Don't make the mistake of assuming that "encryption capabilities" means an algorithm. Their "3D capabilities" are not implemented as a renderer in silicon. Their "multimedia capabilities" do not mean there is a codec on the chip.
Most likely, they will be introducing new instructions to make implementation of encryption simpler/faster.
Of course, that's just what we need. More instructions. Instructions are to processors as features are to software. Selling points, branding, market differentiators. Usually unnecessary.
Kinda makes me nostalgic for the whole RISC vs. CISC debate. We need a whooooole new category for these beasts.
x86 + MMX + Katmai + DES = SCISC (super-complex instruction set computer)?
Or is it VCISC?
"Genius may have its limitations, but stupidity is not thus handicapped." --Elbert Hubbard (1856-1915)
From the rumors (again: rumors) I her, it's primarily going to be random number generation. As people probably know, getting good random numbers from a deterministic machine is really tough. Intel's new chips will provide real random numbers (I imagine through radioactive degredation, but I don't know). This would be an incredible boon to possibly the toughest problem in crypto (next to trust management).
Citizens Against Plate Tectonics
Software encryption is the only way to go, since it takes away the power from other entities.
Imagine if the OS were built into ROM, and you couldn't run anything but the Operating System that comes with the computer..
If you have a compiler and open source, you never need to worry about back doors in a particular encryption algorithm.
... had an article about this very topic this morning and how the Government is making it oh-so-difficult to export such chips. I wish I had read the article in futher depth. The Oregonian website (as painfully slow as it is) hasn't posted the article.
My $0.0000000002 anyway.
Without you I'm one step closer to happiness without violence.
IIRC and if it's what I read about a month or so ago. Not meant for standard encryption duties, only to tie a registered product to a specific processor. Scares the piss out of me -- what if you upgrade your processor? What about dual processor systems? What about just buying a new box and moving all your software over?
--
Infuriate left and right
Um, software developers aren't quite imbecilic enough not to have a move-license-across-chips service; they *know* people upgrade.
Unique CPU IDs are lovely, because it means Intel can have a database of what speed rating they sold each CPU at and you immediately get rid of people relabelling CPUs (since the buyer just checks with Intel what their CPUID was speed-rated at) without having to kill off the overclockers (who check what their CPUID was speed-rated at, then ignore that).
Hardware RNG is good because it removes a very hard-to-analyse source of possible insecurities (which Netscape ran into a few years back) from your crypto algorithms.
not that I care much anyway, this would apply to proprietary s/w, which I don't depend on.
If my computer used hardware based encryption, I would use PGP and the hardware based encryption. It would be preferable to sending it plaintext. And if the government decides it wants to peek in on my email, PGP will make sure their plans fail.
Many people have mentioned that doing this may have some kind of potential backdoor by the government. Which is probably more than likely.
However, my argrument with this is that speed will suffer. Intel processors use CISC technology, 1 instruction per cycle. Adding encryption routines would send the processor though the backlogs of hell.
If Intel was to do encryption, they would have to switch to a RISC system, and then the problem is possibly going to turn into the Y2k problem. Chips that use the same base code, including the encryption routines, and then someday, some major worm comes along and kills the encrypt, and shuts down the processors of the world.
IMHO: Hell no.
-Michael J. Lu
"The little secret that haunts Corporate America...a techonology that won't go away."