I have to say, the Spectre exploit is a beautiful concept, utterly astonishing in its universal applicability across decades of technology. It is the software equivalent of using a stethoscope to crack a safe. There's no way to fix it but buy a new safe.
@DialMforMara I'm reading the paper on spectreattack.com
@photopuck …so, where does that leave people who have to worry about security? If this cache behavior is deep enough to not be patchable via µcode updates, what then? Can you turn off branch prediction in µcode at the expense of performance? Will we just have to buy new processors once the next gen of chips comes out and hope they've fixed enough?
What do we do about Spectre?
I dunno. I can imagine some short-term mitigation strategies -- JIT code fogging, for example to at least mitigate malicious website attacks -- but long term, I have no ideas other than micro-architecture changes. Which basically means new hardware.
There's going to be a lot of data stolen over the next few years, that's for sure.
Fortunately people a lot smarter than me are working on this!
@photopuck Induction-powered electric toothbrush*, permanently switched-on, Araldited** to the inside of the safe door?
* Or induction-powered vibrator, if such a thing there be..
** Yes, yes, "to Araldite" is a verb.. it's in the Oxford English Dictionary.. I'm sure.. oh, near the back somewhere..
@photopuck it's a thing of wonderful and terrible beauty.
We are privileged and cursed to see it exist.
@photopuck And as such I anticipate a campaign of purposefully deafening as many people as possible in case they become safe crackers.
@photopuck a shockingly good analogy for a side channel attack
@photopuck Unfortunately it's not a safe that anybody makes yet. In fact the means to make that metaphorical safe does not exist yet. And there should be concerns that the radical reengineering necessary to create a safe that works as well as the safes we're used to will not themselves have even worse vulnerabilities due to not having a couple decades of progressive hardening engineered in.
@photopuck fwiw the scholarly paper on the attack vectors is surprisingly readable; the first section is, references aside, a better introduction to the problem for laypeople than most of the popular media's attempts to explain it.
@photopuck
Do I want to know what it is?