[Cryptography] LUKS on ATA versus on SSD

[Nicko van Someren]

As ever, I have to agree with Jon.

I would add that the TL;DR of the complexity is that for both SSDs and RRDs (Rotating Rust Drives), for many years it has been the case that when you "write to a block" there is no guarantee that what was in "that block" before will get overwritten any time particularly soon. Both types of drive remap blocks and apply "whitening" to the stored data before writing to the media, but none of this really matters because an attacker is just going to ask the drive for the data and the drive is going to give it them.

Unusable encryption that has incrementally better theoretical properties than fractionally less perfect but usable encryption is less secure in practice. Cryptography may be hard, but key management is harder. As Jon says, take a laptop that has a good user experience and don't worry about it.

Cheers,
Nicko
 

> On Dec 31, 2025, at 20:55, Jon Callas <jon at callas.org> wrote:
> 
> My personal opinion is that this is basically BS. There's enough truth in there that it's hard to pry out the silliness.
> 
> Every drive today is basically running its own operating system. Yes, an ATA disk is just storing zeroes and ones, but it is doing many of the same things an SSD is doing, like remapping dodgy blocks to good ones. The ATA presents an interface to the driver that is just a bunch of blocks, but underneath, it's complex.
> 
> In contrast, an SSD is also incredibly complex underneath and presents to the driver an interface that is just a bunch of logical blocks.
> 
> In each case, we have a situation of OMG, complexity is scary, especially when it's so proprietary that we can't talk coherently about it. In all these, there's so much complexity that what we don't know is much more than what we do.
> 
> Me, I wouldn't worry about it. Take a laptop that has good user experience for you.
> 
> Jon
> 
> _______________________________________________
> The cryptography mailing list
> cryptography at metzdowd.com
> https://www.metzdowd.com/mailman/listinfo/cryptography