Re: [s-t] needle in haystack digest #3 (fwd from Nick.Barnes@pobox.com)
No Such Agency doesn't fab much of anything; they can't afford to. They and their ilk are far more interested in things like FPGAs and adapting numerical algorithms to COTS SIMD hardware, such as graphics processors (a la http://www.gpgpu.org/). Why do they have their own fab plant if they don't fab anything? http://www.globalsecurity.org/intell/facility/nsaspl.htm
On Thursday, November 6, 2003, at 09:20 AM, Dave Howe wrote:
No Such Agency doesn't fab much of anything; they can't afford to. They and their ilk are far more interested in things like FPGAs and adapting numerical algorithms to COTS SIMD hardware, such as graphics processors (a la http://www.gpgpu.org/). Why do they have their own fab plant if they don't fab anything? http://www.globalsecurity.org/intell/facility/nsaspl.htm
The conventional--and convincing to me--story has been that they had National Semi (and maybe others) help them with local fabs. These are fabs for things like key chips (the ICs carrying keying material in whatever form, for Permissive Action Links, and ultra-sensitive kinds of stuff that they wouldn't the usual cranked-up fab workers in Sunnyvale or Nampa getting near). I heard ten years ago that the National Semi fab on-site was a lowly 2-micron fab. Which was enough for keying material. Crunching chips, for special purpose computers, don't carry the same security requirements, as the secret stuff in the code that is being run and not the fuses or links being blown. For this, they would use whatever is out there. --Tim May
Tim May wrote:
On Thursday, November 6, 2003, at 09:20 AM, Dave Howe wrote:
No Such Agency doesn't fab much of anything; they can't afford to. They and their ilk are far more interested in things like FPGAs and adapting numerical algorithms to COTS SIMD hardware, such as graphics processors (a la http://www.gpgpu.org/). Why do they have their own fab plant if they don't fab anything? http://www.globalsecurity.org/intell/facility/nsaspl.htm I heard ten years ago that the National Semi fab on-site was a lowly 2-micron fab. Which was enough for keying material. Hmm. according to the link I found and posted, they *started* at 1-micron and has been tracking its "industry partners" improvements in tech, 0.8 microns up to 1995 then .5 then onwards (with an eventual goal of 0.35, although the piece was written in 1995 so they are probably on copper now too)
Crunching chips, for special purpose computers, don't carry the same security requirements, as the secret stuff in the code that is being run and not the fuses or links being blown. For this, they would use whatever is out there. Non-volitile keying material on-chip requires only standard proms - much cheaper just to buy those off the shelf; for that matter Industry standard "smartcards" usually possess cpu, eaprom program and eaprom data areas on a single chip (and the application would actually prefer some sort of dynamic memory whose contents will vanish if the power is removed from the onboard CPU but we can leave that aside for now - smartcard chips often have that too) Some of that capacity is no doubt used and intended to bridge real or artificial chip droughts (if a manu doesn't want to sell them a given chip, or raises the price drastically because he knows how essential it is to some secure device, the NSA can churn out a few thousand to fill in the gap) but there are advantages to having a completely custom chip - if no attacker could possibly know the layout, command set or capabilities of a chip, that makes his job so much harder (not quite STO - if an attacker has only one or two chips to attack, then every time he gets hit by a trap that removes a crackable device; custom chips can have such things as capacitive test pads (for detection of insulation removal) thin conductive (but visually identical) layers that must maintain continuity, and so forth.)
participants (3)
-
Dave Howe -
Eugen Leitl -
Tim May