Ookhoi <ookhoi@dds.nl>:
> Hi Andreas,
[snip]
> > Note that _real_ secure file deletion is much more difficult than simply
> > writing 0's then 1's over the data (as much as the US Government would
> > like you to believe this). I read a paper that suggested 27 passes of
> > specific byte codes + random data were necessary to even have a hope of
> > having the data be securely deleted.
>
> Hmm, do you have an url about this? How is it possible that data is
> still on disk if you write 0's and 1's over them? (or am I missing
> something?).
I think this is a reference to the mechanical (and magnetic) variances in each
write operation.
1. head positioning is not exact. Overwriting a track/sector only overwrites
where the head is currently located. There is some bleed over on the
edges of each write.
sector
area
------------_________--------------- (disk radius slice
++++ actual recorded area.
With multiple writes the magnetic surface becomes something like
------------_________---------------
4444
3333
2222
1111
Where each overwrite doesn't quite cover the entire sector area.
The final surface has the following:
------------_________---------------
1244443
This allows a scanning magnetic microscope to actually read the older
tracks.
This is sometimes done by the data recovery companies when a client
has a catastrophic failure and absolutly must recover the older data.
2. In addition to the surface scan, magnetic fields penetrate the surface
to different levels. The earlier writes can go deeper since the material
is (presumably) in a more homogeneous state. As a preferred field direction
is set (a write), succeeding writes must overcome that orientation. Since
these fields weaken at the maximum penetration depth, the deepest fields
will tend to not get erased. When this is combined with the mechanical
positioning error, a LOT of data is left resident on the media.
This allows for a very expensive recovery by
a. perform a surface scan using a magnetic microsope
b. CAREFULLY polish off the top layer
c. scan the surface again.
d. repeat until down to metal.
-------------------------------------------------------------------------
Jesse I Pollard, II
Email: pollard@navo.hpc.mil
Any opinions expressed are solely my own.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun Jul 23 2000 - 21:00:12 EST