[ale] destroy old drives

[Jim Kinney]

Thread. Spools. Sewing. Ha!

On April 12, 2019 8:29:04 AM EDT, "Lightner, Jeffrey via Ale" <ale at ale.org> wrote:
>Just as I would a series of emails about sewing spools, I’ve lost
>interest in this thread.  :p
>
>
>From: Ale <ale-bounces at ale.org> On Behalf Of Jim Kinney via Ale
>Sent: Friday, April 12, 2019 8:17 AM
>To: Alex Carver <agcarver+ale at acarver.net>; Atlanta Linux Enthusiasts
><ale at ale.org>
>Subject: Re: [ale] destroy old drives
>
>Heat above Curie point will always remove all data from all recovery
>methods.
>On April 12, 2019 1:22:49 AM EDT, Alex Carver
><agcarver+ale at acarver.net<mailto:agcarver+ale at acarver.net>> wrote:
>
>No, the magnetic strength of the head itself hasn't changed because
>there's still a minimum field required to flip the domain of the
>material which is intrinsic to the material coercivity.  Instead the
>head has changed technology from old wire-wound heads to metamaterials
>that exhibit giant magnetoresistance.  The head is now physically much
>smaller so it can confine the field into a tiny area while avoiding
>some
>of the divergence that can overwrite adjacent tracks.  Remember that
>the
>platter already has servo tracks buried within it which the drive does
>follow to maintain positional tolerance so there's certainly more than
>one layer of material present in the platter.
>
>Even still, print through happens but it does take time.  It's also not
>necessary for the domains below the surface to fully align.  The modern
>hard drive is an exercise in digital signal processing more than using
>a
>hysteretic circuit to detect the individual domains.  The domains are
>so
>small and traveling so fast past the head that it's actually an spread
>spectrum RF signal being transmitted from the read head.  Modern drives
>don't read individual ones and zeros anymore, they read a waveform and
>use statistical processing to recreate the bits on the other end.
>
>The magnetic domain won't be fully erased with a basic rewriting
>program, there will always be a residual field especially if those
>particular domains are not rewritten very often.  So it's not
>impossible
>to do some additional DSP magic and deconvolve what is written on the
>surface from what is deeper in the layer, information that will subtly
>alter the waveform.  We already rely on this technology in things like
>the GPS system and other similar spread-spectrum/ultra-wideband
>devices.
> A determined actor could filter out the strong surface signal and
>eventually recreate a good portion of the underlying signal.
>
>On 2019-04-11 14:09, Jim Kinney wrote:
>
>Really? The only groups that want the data _that_bad_ have subpeonas.
>The other groups that can read around the holes already have your data.
>All you're really trying to do is make sure the drive is not usable for
>the basic computer bad guy.
>As areal densities have increased exponentially from 10M drives to 10TB
>drives in the same space, the size, and thus strength, of the magnetic
>domains has decreased exponentially. So the bleed over has also
>decreased. The transition to vertical magnetic domains  has made the
>crosstalk to the platter substrate nearly zero. Add in the platters are
>simply not magnetizable at all and there's basically no data bits
>anywhere possible except on the platter surface.
>bad ascii art:
>N-S   bit domain on surface___    platter surface
>S-N   induced bit domain subsurface
>  N  S   |    |        2 adjacent vertical domains 1
>0  S   N____       platter surface
>  N - S     Induced data bits are just wrong! Now mix in the 2D spacial
>arrangement and which subsurface pole pairs with which other? No
>monopoles in magnetic media (yet :-)
>
>
>On Thu, 2019-04-11 at 13:31 -0700, Alex Carver via Ale wrote:
>
>If someone really wants your data, holes don't matter.  The rest of
>theplatter is still intact in that case and can have the data
>extracted.
>There's also no guarantee that Dban can write enough to be sure that
>themagnetic domains are fully randomized deep in the platter.  The
>longerdata sits statically on the disk  the more opportunity for the
>surfacedomain to imprint on deeper domains (this is actually a
>problem withmagnetic tape, magnetic data can print through from one
>layer of tape tothe next layer when it's wound on the spindle).
>A serious entity can perform a deep level scan of the platter
>andretrieve the low level signal under the surface domains and see
>previousdata.  The drive head typically isn't powerful enough to
>write thatdeeply because it has to keep the tracks narrow.
>On 2019-04-11 12:13, Steve Litt via Ale wrote:
>
>On Wed, 10 Apr 2019 22:11:42 -0400Jim Kinney
><jim.kinney at gmail.com<mailto:jim.kinney at gmail.com>>
>wrote:
>
>Dban advantage: it can be done across hundreds or thousands of
>drivesbefore larcenous third party "shredders" physically touch
>the drives.
>
> That's a good point.
> Doesn't dban take an hour or more? How many drives can I do with
> onecomputer? How long would it take to test whether each is really
> blank?
> What might be nice with 1000 drives to do is dban followed by
> drilling3 holes in each drive. I'd say each drive would take 1
> minute for 3holes, so it's about 2 days for one employee to drill
> the holes. Or,perhaps, one employee could both dban and drill the
> holes, drilling theholes while the next batch is dbanning.
>
>_______________________________________________Ale mailing
>listAle at ale.org<mailto:listAle at ale.org>
>https://mail.ale.org/mailman/listinfo/ale
>See JOBS, ANNOUNCE and SCHOOLS lists at
>http://mail.ale.org/mailman/listinfo
>
>
>
>--
>Sent from my Android device with K-9 Mail. All tyopes are thumb related
>and reflect authenticity.

-- 
Sent from my Android device with K-9 Mail. All tyopes are thumb related and reflect authenticity.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.ale.org/pipermail/ale/attachments/20190412/530ca010/attachment.html>