Has anybody else seen the recently published exploits Meltdown and Spectre? Here's a link: https://meltdownattack.com/
I'm wondering if this will increase in Risc-V processors, as most will not be vulnerable to this exploit. It relies on speculative and out-of-order execution which most current Risc-V processors do not have.
Louis Pearson desttinghimgame@gmail.com writes:
Has anybody else seen the recently published exploits Meltdown and Spectre? Here's a link: https://meltdownattack.com/
The thing about Meltdown/Spectre is that they're really only problems if you rely on sandboxing to run untrusted code.
This should be more incentive to run on fully free software. If the only code you run on your machine is free software, then there's essentially zero risk of Meltdown/Spectre being an issue. An important point to highlight is that this includes JavaScript that most people run in the browser. The JavaScript Trap [1] as Stallman explained a few years ago.
If people can take back control of their computing but running free software and moving off virtual servers to dedicated serveres or their own product like FreedomBox [2] then issues like meltdown/spectre don't matter.
[1] - https://www.gnu.org/philosophy/po/javascript-trap.ja-en.html [2] - https://freedomboxfoundation.org/
On Thu, Jan 04, 2018 at 06:13:45PM -0500, Adam Van Ymeren wrote:
Louis Pearson desttinghimgame@gmail.com writes:
Has anybody else seen the recently published exploits Meltdown and Spectre? Here's a link: https://meltdownattack.com/
The thing about Meltdown/Spectre is that they're really only problems if you rely on sandboxing to run untrusted code.
It doesn't care whether you sandbox. It makes a privilege escalation possible. If untrustworthy code runs with few privileges, it can exfiltrate enough information to accomplish a privilege escalation. The point of mentioneing the sandbox is simply that the sandbox doesn't help.
Of courses it doesn't matter if you trust the code. It matters if it is trustworthy.
-- hendrik
Hendrik Boom hendrik@topoi.pooq.com writes:
On Thu, Jan 04, 2018 at 06:13:45PM -0500, Adam Van Ymeren wrote:
Louis Pearson desttinghimgame@gmail.com writes:
Has anybody else seen the recently published exploits Meltdown and Spectre? Here's a link: https://meltdownattack.com/
The thing about Meltdown/Spectre is that they're really only problems if you rely on sandboxing to run untrusted code.
It doesn't care whether you sandbox. It makes a privilege escalation possible. If untrustworthy code runs with few privileges, it can exfiltrate enough information to accomplish a privilege escalation. The point of mentioneing the sandbox is simply that the sandbox doesn't help.
Yeah I didn't phrase that quite right. I meant that these vulnerabilites make it impossible to sandbox malicious code.
Of courses it doesn't matter if you trust the code. It matters if it is trustworthy.
Indeed.
On Thu, 4 Jan 2018, Adam Van Ymeren wrote:
The thing about Meltdown/Spectre is that they're really only problems if you rely on sandboxing to run untrusted code.
I'm not convinced that sandboxing is only useful for untrusted code. Sometimes my trusted code has bugs (e.g. I would like to be able to look at random images or documents or expose my webapp to the world), and I would really like for it to not be able to be tricked into doing something it shouldn't. I would also like to be able to compute in shared environments.
Best, Jack
On Thu, 4 Jan 2018, Jack Hill wrote:
On Thu, 4 Jan 2018, Adam Van Ymeren wrote:
The thing about Meltdown/Spectre is that they're really only problems if you rely on sandboxing to run untrusted code.
I'm not convinced that sandboxing is only useful for untrusted code. Sometimes my trusted code has bugs (e.g. I would like to be able to look at random images or documents or expose my webapp to the world), and I would really like for it to not be able to be tricked into doing something it shouldn't. I would also like to be able to compute in shared environments.
Oh, I guess it might be helpful to explain a little bit more why I would like to be able to continue to use shared computing environments. I've been increasing amazed and intimidated by what it takes to understand modern computing. One of my outlets for these emotions is Hcoöp [0], which is an internet service hosting coöperative. We run services such as email, web, and file collaboratively, which saves any one person for having to do all that work on their own. I appreciate the work of projects like Freedombox, but why stop the collaborating after writing the code? I want to be able to collaborate on running it as well!
In addition, some things just don't make sense for all of us to own on our own. I might not often need a large memory or hundreds of core compute cluster, but when I do, it is nice to be able to use a shared resource rather than purchasing my own.
Best, Jack
arm-netbook@lists.phcomp.co.uk