You're still missing the point. There are a lot of hardware vulns out there, I know the majority are not side channels. I'm only talking about side channels here, see: "whole new attack surface". Theres a lot to be said for vendor flaws but that's not what I'm talking about here.
-
-
Replying to @hedgeberg
Assigning blame here is hard. Si vendors knew about side channels but repeatedly disregarded and downplayed the threat, and underfunded security in general. No one will ever know if Spectre could have been prevented, but the industry didn't even try.
1 reply 0 retweets 2 likes -
Replying to @syncsrc @hedgeberg
In fact, they designed entire new security infrastructure that fundamentally depended on the absence of side channels. With potentially catastrophic consequences if it gets widely adopted and broken.
2 replies 0 retweets 5 likes -
SGX aside, nevermind virtualization and the whole cloud craze. Not all side channels are created equal. There's a reason nobody but Intel is vulnerable to L1TF. Nobody in their right mind would think short-circuiting EPT is ever okay, even if you're in speculation.
1 reply 2 retweets 5 likes -
Replying to @marcan42 @matthew_d_green and
I get that the industry wasn't really on the ball on this one, but there's a pretty massive difference between Spectre v1 (a fundamental security issue with speculation that affects the whole industry) and L1TF (Intel what the fuck are you doing?), other bugs being in the middle.
1 reply 0 retweets 5 likes -
Replying to @marcan42 @matthew_d_green and
This sounds like no vendor had a proper security team evaluating these kinds of side channels, but on top of that it seems Intel likely had an explicit policy of "literally anything goes before instructions retire, no matter how insane", which is crazy.
1 reply 1 retweet 5 likes -
This is exactly the damn point though Marcan, when you're fighting with a million other factors and trying to squeeze out performance that's an easy mistake to make, and an easy one for a red team to overlook. It's not /good/, and it's not ok that its a problem that exists...
1 reply 0 retweets 1 like -
Replying to @hedgeberg @marcan42 and
...but this entire attitude is incredibly unhelpful. There are no good security verification tools in silicon design beyond manual review. It's easy for us to call out Intel, but we're post-discovery now. How do we make sure this doesn't happen in the future instead?
2 replies 0 retweets 1 like -
Replying to @hedgeberg @marcan42 and
Pretty sure the answer isn't just "fuck Intel" and it's not "down with CISC", this is a lot more nuanced than any of you are taking the time to acknowledge. This was likely the mistake of a single engineer that noone caught, and that's understandable becsuse...
1 reply 0 retweets 1 like -
Replying to @hedgeberg @marcan42 and
...CPU architecture design students werent taught this stuff until really recently. It's easy for us to call them out, but how are they supposed to know if they weren't taught? Shaming mistakes instead of teaching people how to do better only causes more problems.
2 replies 0 retweets 0 likes
I'm shaming the *company*, not the *people*. Intel has more than enough money that they should've cared about this stuff, they should've hired people who could point at things like this and say it's a bad idea.
-
-
Replying to @marcan42 @hedgeberg and
The problem isn't the people working for Intel, it's that Intel allowed things to get this ridiculous. And usually when things get this ridiculous *someone* notices, and I bet someone did, and I bet someone in management shut them down.
1 reply 0 retweets 1 like -
That's a pretty bold statement to be making considering how little we know about the internal operations, and whether or not you think you're shaming the company, they don't care, the people that do care are the engineers there and everywhere else that read what we say.
1 reply 0 retweets 0 likes - Show replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.