Information security, recycling and irony

[Ben Scott]

On 2/2/06, Tom Buskey <tom at buskey.name> wrote:
> So, to bring it back to topic, how do people go about removing data from
> computers?

  For corporate data and internal re-use of equipment, I generally
just use "Delete".  Regular users don't have the privilege to do
anything more (and if they've cracked that, they have most likely
cracked the server where the master copies are anyway).  Sometimes,
I'll do a "write zeros to all blocks" operation.

  For regular corporate equipment being discarded, sold, or otherwise
going out of our control, I use DBAN.  I actually think a simple zero
would suffice for realistic purposes, but it doesn't cost me anything
more.

  For computer media containing classified information, we don't have
a choice.  We're required to send the media to the NSA, who destroys
it for us.  Apparently, many contractors are not trusted to destroy
computer media themselves these days.

  When classified data destruction is performed, the "secure wipe" is
often disallowed now, too.  So is simple shredding of floppies and
CDs.  The only acceptable method is "obliteration" -- grinding to fine
powder or burning to fine ash.  Information densities have reached the
point where even a tiny fragment can be a big problem.  At the same
time, devices are becoming more intelligent.  Things like cache, spare
sectors, and error correction data mean the old "secure wipe" methods
are not good enough anymore.

  It's actually becoming difficult to build a system (from commodity
parts) for classified processing that can be easily sanitized without
destroying it.  Everything has it's own processor and flash memory
these days.  My mouse has more computrons then an early PC [1].

> Some it also depends on having a badblock scan before data touches it.
> Anyone know how to do this in windows?

  "CHKDSK /R" does a read test of every block on disk.  That's not
always useful, though.

  I used to put a lot of trust in "badblocks -w", but recently I've
had a few disks pass that with flying colors, only to crap out in
production use.  I haven't had a chance to investigate further.  I
suspect it's the same design intelligence problem mentioned above.  It
makes failures less likely to happen, but harder to diagnose when they
do happen.

Footnotes
---------
[1] I made that up, but I'd bet it's true.

-- Ben