03:51
@NicHartley Sorry that's not something you'll be able to do with COTS products.
An HSM requires tamper-resistance among other things, and a regular old microcontroller for the processor is not what you'd want. An HSM has features that go down to the silicon level, to make things like chip-off much more difficult. So what you'd have to do? Find an IC fab who can build it to your specifications. It will be very expensive. Perhaps even millions.
@NicHartley Ok, a few are:
1) Internal sensors to detect changes in ambient EM or temperature
2) Anti-clock and power glitching features
3) A special kind of "glue" in the IC which makes chip-off difficult to do without damaging it.
There's a whole lot more, much of it probably requires domain expertise to find.
Well in this case, usually radio waves of various kinds, or higher-frequency stuff like X-rays.
E.g. A burst of X-rays can "freeze" CMOS RAM in place, which of course is not good.
Well not just X-rays, but other ambient noise or anything that causes an inductive current.
Plus remember that an HSM is passive - it's powered entirely by the (untrusted) device it's hooked into, which means it also has to treat even its power source as untrusted.
Same with the clock source (clock glitching is a common way to bypass 8051 fw locks, etc).
Some do, but not all. Usually a multiplier for an external clock. But even with an onboard clock, it can typically be influenced by power glitching or EM or temperature, so sometimes even an internal clock can't be trusted. And of course they're hyper-vulnerable to laser fault injection.
It also has to be designed with special materials to prevent X-ray inspection.
The thing about HSMs is that they have to fight against the one threat that most threat models need to accept - unconstrained physical access (in a case where "use a lock" isn't enough). I mean, you have to design it so it's secure even if an attacker can flip memory bits with an arbitrary probability or cause internal faults at will, and control the power source and sometimes clock source, ambient temperature and all other ambient conditions, and can remove many layers of protection.
@NicHartley Yeah then you know the kind of difficulty that this can bring. Except in this case, you're not worried about space causing faults, but space trying its hardest to induce faults that then make side-channel attacks possible that otherwise wouldn't be and monitoring with sensitive equipment.
"So let's find a way to induce a fault in the retry counter!" :P
How "from-scratch" are you thinking anyway? Building an HSM from a secure SoC or microcontroller is not nearly as hard as developing the microcontroller and sending the scheme to a fab.
There are a lot of secure SoCs with trusted bootloaders and all that which can be used in a HSM.
Yeah a lot of industry secrets. Hardware security is not my forte (though I find it interesting).
Sadly physical security can't protect from some adversaries (e.g. legally-privileged ones).
That's true, and early detection can certainly work well.
(Though thermite is a little more risky than, ya know, just powering down)
Yeah but then you're going to prison for a long time for illegally handling explosives.
Better to just trust strong encryption.
Yeah but igniting it in a residential home is.
That's probably even worse, since you're putting a lot of businesses at risk.
Still better data destruction techniques though.
Oh in-house... maybe if you get a lot of permits.
I mean it's not like only the military can use it, but it's not as easy as SED.
I'd be more afraid of leftover magnetic remnants than AES or Serpent being cracked.
In theory yes, but it doesn't heat it evenly.
@NicHartley There are existing self-destructive techniques for some chips already.
Literal self-destructing SoCs. You could keep an encryption key in one of those.
But at that point you're already admitting adversaries more powerful than the NSA. :P
So you'd have to take into account things like EMSEC. In that case, I'd probably be comfortable only with a minimum 100dB reduction in all EM and electric currents at the relevant frequencies.
Then you might be part of the problem. :P
Industry pays better anyway. Better than the FBI or RCMP at least.
With two people typing on the same keyboard to defend from hackers!
But can he track a killer's IP address in real time using a Visual BASIC GUI?
Have you ever seen "Kung Fury"? With the hacker who hacks time?
It really is. Standing on what, a C64 or something to hack back in time?
I love when he's typing on the keyboard and he starts typing out random code, then the digits of pi, then just asdfasdfasdf. :D
I mean... I guess he's right... technically.
Well, if your far-future descendants live past the black hole era and heat death.
The New Last Question - How do we prevent heat death^W^W^Wbreak a 128-bit key.. :P
I mean it's possible for a 128-bit key, but a 256-bit key begins to bring the total energy of the universe as a limiting factor, even with perfectly efficient computation.
And 2^256 is a big number.
Now, it's absolutely possible that cryptanalysis will break modern ciphers.
Though I doubt we'll see that made practical any time soon for strong ciphers.
Eh... I recall that there's not enough mass in the visible universe to do that.
I think you could just barely count to 2^128 with the entire energy output of the sun from birth to conversion to a black dwarf, and 2^256 is far, far higher.
