cisin at xenosoft.com
Thu Jan 7 11:36:31 CST 2016
>> 1) if the alignment of the head of the original recording and of the
>> overwrite head are not a perfect match, then there can be some residual
>> data somewhat off axis.
On Thu, 7 Jan 2016, Christian Corti wrote:
> At a first thought I don't see how there can be residual data because there
> is the tunnel erase head after the R/W head. The drives must be very
> misaligned (i.e. more than the width of one erase half) to still have
> residual data.
Yes, that makes sense. But, the erase head probably doesn't scramble the
weak residual flux transitions as much as overwriting does.
>> 2) if the data was overwritten once, with a known pattern, then somebody
>> with sufficient resources and motivation can attempt to analyze the noise,
>> and determine "what, overwritten by a 0 could produce the noise that we
>> have here." Accordingly, there are guvmint standards of MULTIPLE patterns
> That is why you don't take /dev/zero but /dev/[u]random for overwriting data.
Even then, since the s'posedly random data that was used for overwrite is
still readable, leading to, "OK, here's what's currently on it, what prior
recording would generate THIS background noise?", thus, a scond overwrite
of a different pattern would render it past any "reasonable" efforts.
I've heard that there are "standards" for a number of overwrites, and what
patterns to use, . . .
But, as I mentioned before, the most thorough protection of all, is to be
too boring for it to be worth any effort at all.
'Course extremely violent total destruction of the drive has its own
I suspect that most of the stuff that impresses the hell out of outsiders
to the field is really nothing more than patching DIRectories to "UNERASE"
files that haven't even been overwritten.
More information about the cctalk