That's why all the FBI "we're going dark!" bullshit is so obviously just that: bullshit.
People are such idiots that end-to-end encryption is not hurting them.
I mean ffs most pedos even use non-anonymous P2P technology.
Enough to keep the FBI occupied for decades.
@J-- Dunno. He talks a lot about it on Twitter. My guess is that he started working on them back before LLVM had good plugins (and back before the Linux kernel could be compiled with Clang anyway).
And honestly, GCC really needs some love when it comes to plugins.
I'd love to see GCC get a CFI implementation (besides RAP, since RAP, although it's the best one out there, requires too many changes to source code to be practical for anything but the kernel).
@J-- Yeah it's not going to be upstreamed, even though it's really, really good.
However I think Kees is trying to get the kernel ready for forward-edge CFI.
Especially since Windows has had that for years, and now they even have forward and backward-edge CFI.
@J-- It's not something that can be implemented in hardware. What the CPUs are doing is adding support for hardware shadow stacks, which simply makes backward-edge CFI easier to implement.
I've heard of that but I'm not really familiar with it.
> Put simply, when the feature is enabled on a chip, a tag is assigned to each memory allocation, and if that memory is to be accessed it must be made via a pointer with the correct tag. If not, it is reported to the process and it will quit or fail.
> The tags are only 16-bit. At best it can provide a 14/15 chance of catching generic memory corruption bugs. Can do better and provide deterministic protection against a subset of bugs such as all sequential overflows and with some effort use-after-free within N allocations.
Daniel Micay knows his shit.
Yeah, and hardened_malloc.
Exploit mitigation is a fascinating and very complex topic.
We have C to thank for that.
It's also quite depressing.
Every time I dig deep into security mitigations I just feel down.
Because everything is fucked and always will be.
The thing is, I don't really care about anything but my own personal security. Unfortunately, others' insecurity will affect me, since I use their services, I compile source code stored on their servers, etc.
I use Tor and most Tor relays are running on Debian, etc etc.
A security auditor for our servers has demanded the following within two weeks:
A list of current usernames and plain-text passwords for all user accounts on all servers
A list of all password changes for the past six months, again in plain-text
A list of "every file added to the server from re...
