On Sun, Sep 6, 2009 at 5:52 PM, Christopher Walters <cjw20...@comcast.net>wrote:

>  Lee Winter wrote:
>
> On Sun, Sep 6, 2009 at 3:42 PM, Sven Joachim <svenj...@gmx.de> wrote:
>
>>  [snip]
>
>  Jumping into that discussion, here is evidence that this is not possible
>> with modern drives:
>>
>>
>> http://www.h-online.com/news/Secure-deletion-a-single-overwrite-will-do-it--/112432
>
>
> No, that it not evidence.  It is an opinion; possibly a very informed
> opinion.  But security issues often require a skeptical perspective.  In
> this case an expert's statement that he does not know how to retrieve info
> from a drive is abolutely worthless in determining whether anyone else knows
> how to retrieve info from a drive.
>
> [snip]
>
>  That will work up to the value of the information being secured.  But
> once the value of the information reaches an upper limit then it becomes
> worthwhile for people to use more sophisticated techniques, and overwriting
> with a constant pattern becomes worthless.
>
> There is a recently revised NIST standard for securing information.  It
> says very little -- propably because the US givernment has an interest in
> lowering other entities security.  The previous versions of that standard
> were a lot more informative and useful.
>
> BTW, no sensible person ever said that 35 passes were necessary and/or
> useful.  A well-informed and well-intentioned expert answered a silly
> question and his answer boils down to the (valid) claim that it is not
> possible for any drive to require more than 35 passes.  The total of 35 was
> obtained by summing all of the possible overwrite techniques for all
> possible drive/recording technologies.  After that many non-sensible people
> claimed that 35 passes was the ne-plus-ultra in disk scribbing, which claim
> is both invalid and stupid.
>
> Lee Winter
> NP Engineering
> Nashua, New Hampshire
>
>
> Not a fan of Peter Guttman, I take it.
>

Incorrect.


>   He is pretty well known in the fields of computer security and data
> deletion.  Here is a link to his paper.
>
> http://www.cs.auckland.ac.nz/~pgut001/pubs/secure_del.html<http://www.cs.auckland.ac.nz/%7Epgut001/pubs/secure_del.html>
>
>
Yup, that's the one.

Please read the first paragraph of the section entitled "Epilogue" and
compare it to the summary I gave above in the last paragraph.  It was my
intention that they mean the same thing.

In the section entitled "Further Epilogue" he goes on to describe the
hopelessness of trying to recover info from a modern drive.  That section is
the target of my comments re the opinion of an expert who states he does not
know how to accomplish a certain task.  There is absolutely nothing wrong
with his statement.  There is everything wrong with the reader
mis-interpreting his statement as evidence or proof that the certain task is
infeasible.

Case in point.  A couple of years ago one of the major financial companies
(3rd I think) in NYC was concerned about reducing costs.  They run an
IT-intensive operation so they recycle machines often.  That is expensive.
But a 1-or-2-year-old HD is reasonably valuable.  So they wanted to recycle
them rather than destroy them.

But they have _extremely_ valuable information on even their desktop
drives.  Many contain customer information, so, as a fiduciary, any
preventable leakage would essentially put them out of business.  Other
drives may contain strategic information either in the form of documents or
in transaction records. And of course the data-center drives are even more
valuable.  What's all that information worth?  Many zeros.  _Many_.

As an aside, shredding is a popular method of drive destruction.  But modern
drive densities are so high that even a shred 0.01" square can hold valuable
data, so physical destruction alone is not sufficient.  That may have
something to do with the fact that _internal_ gov't standards allow only a
very few kinds of physical destruction -- i.e., complete to the level of a
minimum-sized magnetic domain of the particular recording media.  Think
acid, thermite, grinding/abrasion, etc.

Given the incredible value of the financial system data, how much is it
worth to recover it?  The modern recovery process usually has two distinct
phases, one quite capital- and skill-intensive and one quite ordinary.  The
first phase is to build a data recovery capability (lab).  That takes time,
money, and skilled labor.  The second phase is operating the recovery lab,
which is fairly cheap.  It isn't very fast though (meaning long latency, but
not meaning low throughput).

Point is that once you have such a capability many unreasonable
possibilities become quite reasonable.  And industrial espionage is a
thriving industry.  Just who owns (or more importantly controls) the
recycling company that hauls away your machines/drives?

Have you ever taken a drive apart, replaced the drive electronics and resold
it?  People do it all the time.  And if the replacement happens to be a much
more sophisticated board, you can read lots of things that the original
drive electronics could not.  If you open the sealed module to get at
(replace) the HDA you can do a great deal more.  If you have serious cash
available, like $1e8, you can get a low-level image of the platters with an
STM and turn a computing cluster loose on the image.  That's just an
(extremely) advanced form of OCR.  C.f. "palimpsest".

Point of this mini-rant is that it is very easy to underestimate security
threats.  People should resist that tendency.

Disclaimer: I no longer participate in drive recycling, so I'm not "selling"
anything here.  Just providing a word to the wise.

Lee Winter
NP Engineering
Nashua, New Hampshire

Reply via email to