Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 11 Oct 2024 10:34:10 -0500
From:      robert@rrbrussell.com
To:        questions@freebsd.org
Subject:   Re: How to zero a failing disk drive before disposal?
Message-ID:  <746da1d5-72f8-42fc-917b-f4de0e4d98f9@app.fastmail.com>
In-Reply-To: <2544410a-8a99-4b2e-a194-c8326a2e0ddd@heuristicsystems.com.au>
References:  <5117.1728561469@segfault.tristatelogic.com> <2544410a-8a99-4b2e-a194-c8326a2e0ddd@heuristicsystems.com.au>

next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, Oct 10, 2024, at 21:42, Dewayne Geraghty wrote:
> A good question Ronald.  I worked for a provider of services for the=20
> statutory care of children (eg removed from parents). There are=20
> significant penalties for certain types of information loss.  We=20
> bench-drilled the hard-disks before sending them (out of our chain of=20
> custody) to a furnace.  Admittedly this is an extreme case and for the=20
> reasons already stated in this thread, there was no other way to ensur=
e,=20
> say a name and location, were not available.
>
> And yes, all machines have full disk encryption (FDE).
>

I agree with securing the data, but drilling is not thorough enough to s=
top recovery and marks the drive as a high value target. You are basical=
ly betting that microscopic magnetic domain scanning will not improve fa=
ster than the areal storage density of the drives you throw away. Not a =
bet I want to take personally.

>
> For personal devices we overwrite the device multiple times, though I'=
m=20
> interested in what a "ATA Secure Erase" does to a healthy storage devi=
ce=20
> and whether all sectors are touched?

ATA Secure Erase is a firmware level overwrite and reread of all physica=
l drive sectors. I have seen the drive=E2=80=99s bad sector count increa=
sed by this process. It takes slightly longer than a simple dd if=3D/dev=
/zero of=3Ddrive.

I don=E2=80=99t suggest filling a drive entirely with zeros or ones. Tho=
se writes can be =E2=80=9Coptimized=E2=80=9D out by the firmware. It mig=
ht set a bit flag in the sector header instead of performing the full wr=
ite. They don=E2=80=99t complete faster, they just aren=E2=80=99t guaran=
teed to actually overwrite the old data. The why has to do with a very l=
ong side tangent about how MFM encodings with error correction work.

If you want to overwrite use anything except for a string of all ones or=
 zeros.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?746da1d5-72f8-42fc-917b-f4de0e4d98f9>