3 ms·

Programmable Chips Secured by Chaos

41 points by adrian_mrd 6y ago

rubicks 6y ago
Is this the technology that https://dustidentity.com/ https://dustidentity.com/ is doing?
- paulgerhardt 6y ago
  No. Dust Identiy is basically just making "glitter seals" which have been around for forever (casual search: with a shoutout from HN's own @rdl: https://lifehacker.com/use-glitter-nail-polish-to-make-your-laptop-tamper-proo-1493599646 https://lifehacker.com/use-glitter-nail-polish-to-make-your-...) - while I haven't looked at their patent I wouldn't invest personally given a ton of prior art. PUF's are a completely different mechanism and have also been around for since at least 2001. The most basic version is: every chip has slightly different performance when benchmarked at a fine enough level - your access times for RAM will vary by picoseconds from one chip to the next - you can use those delta's to fingerprint chips and use those fingerprints to seed keys. What follows is 20 years of research improving the technology readyness level of that idea so chips can sign things reliably regardless if they're operating at 60ºC or 10ºC - on power cycle 1 or power cycle 20,000 but uniquely enough so one can't emulate them through trial and error.
shirleyquirk 6y ago
full tin-foil hat time: knowing that state-level actors work with manufacturers to include 'backdoors' into hardware (thinking of RDRAND) is this an auditable source of randomness, or just as opaque?
- astrange 6y ago
  > knowing that state-level actors work with manufacturers to include 'backdoors' into hardware (thinking of RDRAND) We certainly don't know that for rdrand, and not that often for other things either.
motohagiography 6y ago
I've seen PUF's come up a few times lately, and I'm bearish on them, so YMMV on this, however here's why. I'm a bit bothered that the article uses "hackers," as their threat actor, because the only meaningful threat actor in the design of new cryptographic primitives is a group of hundreds of other professional mathematicians and cryptologists employed by an adversary nation state. From purely a security architecture perspective (and not a working cryptographers perspective), they would need a way to be formally verified and testable, which hiding an entropy source in a sealed unit is not. If you want to do any serious applications, you need to go through NSA test labs, which as far as I know, these have not yet and it's a pretty high bar. Given they are physically unclonable, it doesn't provide any additional flexibility for updating key material, which means they are now physically unreplacable devices. That "it's just so random," only means it is random over a field of a certain size, and even then, I've seen "cryptographic" systems where the period of the sample (or sample size) from a random source was reduced to only a few usable bytes right after it was sampled. The entire history of crypto backdoors is about reducing the entropy of seeds and other primitives, so that these PUFs happen to produce noise doesn't change that. Further, the only reason anyone adopts crypto is because it is certified by NIST, and because institutions and certification bodies are like /dev/null for compliance risk, so the decisions are not based on the ability of users to reason about the crypto implementation. The source of risk in security right now are not in P/CRNGs so much as the orthogonal problem of verifying the integrity of implementations of cryptosystems that are well understood, a need which PUFs appear (to me) to specifically frustrate. DeFi has created more of a wild west than a renaissance in cryptography right now, and PUFs seem a bit predatory to me. I know there are some a-list crypto evaluators here, and I'd be interested in how these perceptions of mine are wrong.

41 points by adrian_mrd 6y ago

rubicks 6y ago

Is this the technology that https://dustidentity.com/ https://dustidentity.com/ is doing?

paulgerhardt 6y ago
No. Dust Identiy is basically just making "glitter seals" which have been around for forever (casual search: with a shoutout from HN's own @rdl: https://lifehacker.com/use-glitter-nail-polish-to-make-your-laptop-tamper-proo-1493599646 https://lifehacker.com/use-glitter-nail-polish-to-make-your-...) - while I haven't looked at their patent I wouldn't invest personally given a ton of prior art. PUF's are a completely different mechanism and have also been around for since at least 2001. The most basic version is: every chip has slightly different performance when benchmarked at a fine enough level - your access times for RAM will vary by picoseconds from one chip to the next - you can use those delta's to fingerprint chips and use those fingerprints to seed keys. What follows is 20 years of research improving the technology readyness level of that idea so chips can sign things reliably regardless if they're operating at 60ºC or 10ºC - on power cycle 1 or power cycle 20,000 but uniquely enough so one can't emulate them through trial and error.

shirleyquirk 6y ago

full tin-foil hat time: knowing that state-level actors work with manufacturers to include 'backdoors' into hardware (thinking of RDRAND) is this an auditable source of randomness, or just as opaque?

astrange 6y ago
> knowing that state-level actors work with manufacturers to include 'backdoors' into hardware (thinking of RDRAND) We certainly don't know that for rdrand, and not that often for other things either.

motohagiography 6y ago

I've seen PUF's come up a few times lately, and I'm bearish on them, so YMMV on this, however here's why. I'm a bit bothered that the article uses "hackers," as their threat actor, because the only meaningful threat actor in the design of new cryptographic primitives is a group of hundreds of other professional mathematicians and cryptologists employed by an adversary nation state. From purely a security architecture perspective (and not a working cryptographers perspective), they would need a way to be formally verified and testable, which hiding an entropy source in a sealed unit is not. If you want to do any serious applications, you need to go through NSA test labs, which as far as I know, these have not yet and it's a pretty high bar. Given they are physically unclonable, it doesn't provide any additional flexibility for updating key material, which means they are now physically unreplacable devices. That "it's just so random," only means it is random over a field of a certain size, and even then, I've seen "cryptographic" systems where the period of the sample (or sample size) from a random source was reduced to only a few usable bytes right after it was sampled. The entire history of crypto backdoors is about reducing the entropy of seeds and other primitives, so that these PUFs happen to produce noise doesn't change that. Further, the only reason anyone adopts crypto is because it is certified by NIST, and because institutions and certification bodies are like /dev/null for compliance risk, so the decisions are not based on the ability of users to reason about the crypto implementation. The source of risk in security right now are not in P/CRNGs so much as the orthogonal problem of verifying the integrity of implementations of cryptosystems that are well understood, a need which PUFs appear (to me) to specifically frustrate. DeFi has created more of a wild west than a renaissance in cryptography right now, and PUFs seem a bit predatory to me. I know there are some a-list crypto evaluators here, and I'd be interested in how these perceptions of mine are wrong.