Writing over a drive using /dev/zero

Christopher Browne cbbrowne-Re5JQEeQqe8AvxtiuMwx3w at public.gmane.org
Tue Nov 30 03:41:58 UTC 2010


On Mon, Nov 29, 2010 at 8:53 PM, Walter Dnes <waltdnes-SLHPyeZ9y/tg9hUCZPvPmw at public.gmane.org> wrote:
> On Mon, Nov 29, 2010 at 11:55:16AM -0500, Daniel Wayne Armstrong wrote
>
>> Any thoughts? Thanks!
>
>  The only thing I have to add is that if you're trying to securely wipe
> a drive, you should use /dev/urandom instead of /dev/zero.  And if CSIS
> or CIA really want the data, they might still be able to retreive most
> of it.  If you really, really want to guarantee the data won't be
> retreived, you have to take out the platters and disolve them in a vat
> of acid.  If the data on the disk isn't *THAT* sensitive, 3 or 4 passes
> with /dev/urandom will usually do the trick.

You're quite right that /dev/urandom (or similar - anything that's
generating random-ish values) should be pretty good at securely wiping
things.

It's rather less obvious that it's possible to realistically get "most
of the data."

Perhaps the NSA has a machine that can en-masse do the kind of
differential analysis needed to draw data off an apparently-erased
disk.  They'd need a special machine that's not available to the
commercial disk recovery industry.

Yes, an examination under special instruments could get bits off, even
after attempted erasure; the trouble is that modern drives have
several complications:
a) Error correcting encodings lead to there being massively different
mappings between where the data physically is and what's being
reported to filesystems and applications, and the mappings mayn't be
evident.
b) Known methods aren't notably fast, and require multiple reads in
order to get the variations in magnetic charges needed to get around
the erasure.  It's liable to take weeks to get data off last
generation disks, and worse for terabyte disks.
c) Machinery for this will be super-expensive, because, much like the
way tape drives are expensive, anything that's not being widely
commercially used, but which, rather, is custom, is high-priced.  This
won't help encourage high performance...

It's *conceivable* that there could be some super-secret NSA machine
to do the job, but there would be *massive* commercial value in making
this available to the commercial data recovery industry, quite likely
more than it's worth to keep the technology secret.
-- 
http://linuxfinances.info/info/linuxdistributions.html
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists





More information about the Legacy mailing list