[118420] in Cypherpunks
RE: Steganos - Wiping Data
daemon@ATHENA.MIT.EDU (Trei, Peter)
Mon Sep 27 11:20:52 1999
Message-ID: <D104150098E6D111B7830000F8D90AE8E62A8B@exna02.securitydynamics.com>
From: "Trei, Peter" <ptrei@rsasecurity.com>
To: "'cypherpunks@algebra.com'" <cypherpunks@algebra.com>
Date: Mon, 27 Sep 1999 11:03:51 -0400
MIME-Version: 1.0
Content-Type: text/plain
Reply-To: "Trei, Peter" <ptrei@rsasecurity.com>
> ----------
> From: Tim May[SMTP:tcmay@got.net]
> Reply To: Tim May
> Sent: Friday, September 24, 1999 10:02 PM
> To: cypherpunks@Algebra.COM
> Subject: Re: Steganos - Wiping Data
>
> At 5:13 PM -0700 1999-09-24, Jonathan Stafford wrote:
> >Less effective than sandblasting, but couldn't you also try wiping the
> >tracks bordering the incriminating track?
> >
> >
>
> We used to have this "how to destroy a disk" discussion every few months.
> Time for someone to ask about thermite.
> The more things change....
> --Tim May
>
>
Yep - this is one of our regular topics. I expect that the
best answer is to *always* use some form of encrypted
sector SW, so the cleartext is never on the disk (and think
hard about swap and temporary files).
Peter Guttmann's paper is a good eye-opener, but I've heard
that even Peter feels that a heavily-overwritten (much more
than 7 times) disk is probably safe against most opponents.
A single overwrite should, IMHO, protect a target against
casual and surreptitious scans on most HDs. The logic
behind this argument is that if it were economic to make
HDs look at the parallel 'overwritten' tracks, HD
manufacturers would use this capability to increase
the capacity of their disks.
Thus, I suspect that reading overwritten tracks requires,
at the minimum, remounting the platters in special drives
with non-standard heads and positioning systems. At the
high end, an opponent could use an Atomic Force Microscope
to examine the media.
An attacker in a black-bag job might be able to sector
copy the non-overwritten data onto a new, identical disk,
swap that for the old disk with 'interesting' overwritten
data, and hope the victim does not notice the substitution.
However, a disk is not a perfect palimpest. Each overwrite
will probabalistically destroy some portion of the previous
data. Even if 90% where recoverable after a single
overwrite, less than 3% of the data is left after 36
overwrites. If sector encryption with a non self healing
chaining technique were used, the first missed bit would
make the rest of the sector irrecoverable, even if the
keys were known.
Just mumbling about the obvious...
Peter Trei
