On Mon, Jun 04, 2001 at 04:30:29PM -0700, Michael K. Sanders wrote:
> Right, you could add random/volatile keys to a block device cipher,
> but doing it at the swap level gives you more control over how long
> keys for pages that are no longer referenced stay around.
> 
> The approach described in the paper divides swap into some number 
> of configurable size sections, with a random key generated on demand
> for each section.  Each key also has associated with it a reference
> count and an expiration time.
> 
> With a block device cipher, all they key(s) would have to persist as
> long as the system is running, no?

As interesting as these methods may be, I remain unclear on how an
encrypted swap with an unencrypted file system does you any real
good.

I mean, except in the circumstance that you have data stored in
memory but never on disk that you wouldn't want someone who
physically stole a drive from your machine, they won't need to dig
through swap, they can just look at the file on disk. (And how often
is information important enough to encrypt not important enough to
store on disk somewhere? Is this *just* to take care of unlocked
private keys? C'mon...)

A cipher pseudo block device clearly makes more sense to solve this
"problem". (I'd argue that if someone has console access, encrypting
your file system isn't going to do much good anyhow.)

Or were you hoping for some other kind of protection from this?
Perhaps for swap over NFS? Why not just encrypt NFS then and get a
more general benefit there too?

       ~ g r @ eclipsed.net