Subject: Re: Cryptographic file storage
To: Ignatios Souvatzis <email@example.com>
From: Dwight Tuinstra <firstname.lastname@example.org>
Date: 02/21/2001 13:10:33
Ignatios Souvatzis wrote:
> On Tue, Feb 20, 2001 at 09:10:14AM +0000, Mark White wrote:
> > ... and it's difficult to get rid of
> > the unencrypted copies from the disk when you want to make
> > is secure again.
> and impossible in the case of LFS.
Not necessarily. The cleaner is in userland, and could be
modified to place such blocks at top priority to be cleaned,
and if not cleaned soon enough, to spawn a process to overwrite
the remnants with random bits.
For that matter, there could be a non-cleaner daemon (called
perhaps `noise') whose job it would be to find and noise out
defunct segments that had been marked for such treatment. This
would allow the cleaner to continue to use the standard cost-
benefit criteria to determine which segments to clean (avoiding
the performance penalty likely to occur if encrypted blocks are
given high cleaning priority just because they're encrypted).
This would mess up the checksums within the segment, so this
would have to be thought about. Also, there would probably
have to be kernel modifications to support tracking which
data blocks need to be noised-out.