Re: "How many overwrites for secure erase?"

I didn't see any mention of the fact that for most disks there
isn't a way to erase revectored (i.e., replaced) blocks.

I looked at "Serial ATA Specification V2.6",
dated 15-February-2007, file name:
SerialATA_Revision_2_6_Gold.pdf
(Available to members or for sale at:
https://www.sata-io.org/secure/spec_download.asp)
and found that the operation of the ERASE and ERASE UNIT commands
were not specified, so, in the absence of a manufacturer's
specification for a particular disk model, one can't be sure
that anything in a sector that has been marked defective is
touched at all.


On Sat, 23 Feb 2008 13:02:40 -0500, Arthur T. <arthur@munged.invalid>
wrote:

> In Message-ID:<Xns9A4DA577A884FYAsfKJXSTO@194.117.143.37>,
> bealoid <signup@bealoid.co.uk> wrote:
>
> >1) The theory is that you don't know what tech your attacker has, and you
> >don't know what tech your attacker will invent in the future, and so you
> >over-write many times with patterns and random data, then take the
> >platters out and physically destroy them.
>
> Pointing out the possibilities of future tech (and the
> near-impossibility of ruling out what future tech might be) puts
> overwriting into a different perspective. Thank you.
>
> >2) You have sensitive information (patient medical stuff, for example)
> >and it's just easier to do the belt-and-braces destroy thing than a
> >sensible destroy, if only to keep the wing-nuts out of your hair. You've
> >removed any doubt.
>
> CYA is a very good reason, but not a technical one ;-).
>
> >People might prefer to do cost-benefit risk analyses - it takes time (and
> >thus money) to overwrite disks.
>
> And, it takes even more time and money to do the analysis of
> how much overwriting is necessary. Thus, we're likelier to get
> "guidelines" than reasoned, technical answers.
>
> So, even if today N overwrites makes a disk unreadable,
> tomorrow someone might find a way to read it. (And, of course,
> even N overwrites might be readable by a closed-mouthed government
> agency.)
>
> Short of a theoretical proof (which unlikely to have much to
> do with real-world technology), N can be argued but might never be
> enough for absolute security.
>
> I think this explains the lack of reasons for the guidelines
> currently available.
>
> Thanks to all who responded to this thread. I responded to
> this post because that's when the answers sunk in, but all of the
> responses were helpful in bringing me to the state where I could
> understand the gestalt.