"Sebastian G." <seppi@seppig.de> wrote in
news:62qjtgF2484t3U3@mid.dfncis.de:

> bealoid wrote:
>
>> "Sebastian G." <seppi@seppig.de> wrote in
>> news:62q4cpF2486s5U1@mid.dfncis.de:
>>
>> [snip]
>>
>>> This is nothing special though, even Peter Gutmann mentions much more
>>> potent ways like magnetic force scanning tunneling microscopy.
>>
>> and even gutmann says that no longer works and that 3 random overwrites
>> is about as good as you can do.
>
>
> And even I wrote that one overwrite is enough, but this doesn't change
> the result that such techniques allow to retrieve information
> statistically significantly better than random guessing. If the high
> level data are encoded with redundancy as well, this might be a problem.

remember that you are allowed as many read passes as you want on each
track. This allows 'averaging out' much of the noise.

You can use heads that are much narrower than the normal read head.

This allows an increased signal to noise ratio for the buried signal near
the outside of the overwrite band since the bands will not overlap exactly.

By using many reads from many 'lanes' through the data, you have a chance
of 'digging' the signal out of the noise.

I think I remember that, at one time, the CIA could dig down through 7
layers of overwriten data to recover the underlying information.

Of course, that may take millions of passes per track.

Knowing what the data could be [a 1 or a 0] makes it much easier than
recovery of an analogue signal (nixon's missing minutes).


--
bz

please pardon my infinite ignorance, the set-of-things-I-do-not-know is an
infinite set.

bz+spr@ch100-5.chem.lsu.edu remove ch100-5 to avoid spam trap