And important part of defense is carefully considering which trust boundaries are defendable and worth defending.
Prikaži ovu nit
At the risk of sounding like a heretic: Given the total cost of Spectre/Meltdown fixes, it would be interesting to do careful threat modelling to decide when and where they are not needed.
-
-
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Being exploitable by running JS makes the set of machines that might need full mitigations very large.. do you know whether current JS engines are still considered vulnerable if their hardware has the costly mitigations disabled? I recall something about disabling highres timers.
-
i am mostly thinking about on-prem servers doing number crunching on non-sensitive data; browsers/clients is a different story...
- Još 9 drugih odgovora
Novi razgovor -
-
-
also i see too many folks proposing "use AMD" as a solution. there's almost certainly similar, unpublished bugs in most major chipsets. i worry sometimes that every time a big bug / leak comes out, folks give out bad advice assuming public knowledge = all possible threats
-
Yet these side-channel exploits are being applied to AMD processors as well, and most don't work or are much more difficult on AMD. This is for a vuln category that is supposedly applicable on most ISAs.
- Još 3 druga odgovora
Novi razgovor -
-
-
That's not heresy: why do I need to take a 35%¹ compulsory hit on my BurpSuite VM running on a machine to which I have exclusive access?
@osxreverser was making the same point too: risk-based mitigations not blanket. __ ¹ more now that the JCC patches will be needed killing VMs? -
For those who still control their own hardware, these occasionally come in very handy for jobs that must be isolated.pic.twitter.com/okOU6RlK1r
Kraj razgovora
Novi razgovor -
-
-
"Just don't mix VMs with different security contexts on the same physical hardware" would be my 80% fix.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
Tweet je nedostupan.
-
-
What I’ve been saying for maybe a year now. You don’t get security boundaries as an implicit consequence of coherent memory access or logical constraints. You don’t get to swap security domains a million times a second. Not with nondeterministic time. Boundaries go explicit.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.