shredding files on a flash drive

[Peter P.]

Kristian Erik Hermansen <kristian.hermansen at ...> writes:
> I won't reply to this any further.  I made my point.  You still didn't

There exists software and there are systems that are able to stitch back
together essentially documents torn to confetti and read what was on them. They
are used by archaeologists among others. The original application was to stitch
back together old documents (even with missing parts!) and tapestries. At least
one application is open source, and serves to digitize old 8 and 16 mm films
using a flatbed scanner. The software stitches the succesive exposures of
several frames each together, picking them out of the background and the film
frame, and makes a modern computer movie from them. 

All modern tape and disk drives do not use alignment between the platters for
any special purpose. Each head has its own data channel. Not working like that
would require the platters and the heads not to have any vibration (including
rotational) to the tune of one half a bit space on the disk. That would be in
the few tens of nanometers at most today. Move this to four or six faces to keep
in sync and a price tag usually well below $300 and you have a picture of
something that cannot be done. $300 won't buy you the motor and control system
that guarantees rotational noise below say 50 nanometers at 7200 rpm no matter
what. Compare to video recording where the heads move against the tape with
something like 2 m/sec and rotational vibration effects are of the order of
magnitude of 10 usec per head rotation at a bandwidth of only 1.5 MHz when all
is well (rarely). A hard disk head channel will have something like 10-20 m/sec
head speed (7200 rpm) and probably 10-100MHz bandwidth.

Any instrument maker will tell you that such things do not work outside a lab.
Tapes have the same problem (longitudinal skew and warping) and deal with it the
same way: each head has its own data channel, clock and data recovery state
machine (including ML masking/decoding), and mini buffer. This has been true
since at least the 1970s when 7, 8 and 9-track tapes ruled the world. Also most
normal hard disks do not stripe the heads and use one head at any one time. Only
high end drives stripe the platters and use all the heads together (but still
with each having its own decoder and mini buffer).

Also there will always be files with known data on the remains (at least sector
headers) that will allow the interleave assembly scheme to be divined. Now add
1+1 together with the confetti assembler above and you have a working recovery
system. Yes, it would take a few weeks of hard work at least. Yes, there are
people who would do such things to get at other countries or companies goodies.
Consider that at current data densities a 1mm length of track could allow the
recovery of 10 kilobytes of data. How many names and addresses or social
security numbers can be stored in 10 kilobytes ?

Imho it is pretty lame to 'make your point' by quoting a paper written by
(knowledgeable) people relatively out of context when there are tons of counter
examples on the net. Sorry for writing this. Add to this that if I know this as
a 'civilian' who is not in the security field at all, there must be people who
know A LOT more. And no, I did not 'prove my point'. I just brought some
arguments into the discussion. This is not a pissing contest as far as I know.

Peter P.


--
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