[Cryptography] In the latest unexpected ransomware twist ...

Jon Callas jon at callas.org
Tue Jun 15 21:24:40 EDT 2021



> On Jun 13, 2021, at 22:48, Phillip Hallam-Baker <phill at hallambaker.com> wrote:
> 
> The problem though is that all three major platforms are trying to compete on security. And that is always a mistake. If keystores are going to be reliably useful, there has to be a cross platform common interface. Preferably arrived at by leveling up, not down.

I disagree a bit. All three major operating systems realize that security is vital. They're all taking different approaches, and it's nice to see that there's creativity there. I just wouldn't call it competition, because no one would switch from one to the other based on the security considerations. Even among the likes of us, it would be a newsworthy thing to switch for that reason, and says more about the person than the technology.

I am hoping that we'll end up with something like CHERI in future hardware. The model we've been using for the last 40+ years of per-process address spaces and VM backed with TLBs and the like is getting a bit teetery to me.

My fantasy is that a 64-bit (or be still my heart, 128-bit) flat address space with capabilities will be *faster* than traditional per-process TLBs stacked ever higher. Even if it's a slight step backwards on the performance end, I think it's worth it. In early virtual memory days, I knew lots of people who bemoaned the performance hit you took with having virtual memory. Now, I'd take a slight performance hit if it meant that I was getting an ASLR analogue with a higher work factor (the reason for 128-bit pointers), mitigations for a variety of timing side channels like Spectre, and so on.

I don't think we need the few options we have on major choices collapsing in to even fewer. We don't need a thousand flowers blooming, but five or ten would be nice.

	Jon




More information about the cryptography mailing list