On Feb 11, 2008 8:11 PM, Jamie Wilkinson <[EMAIL PROTECTED]> wrote:
> Do things like this really exist??  Well, I imagine Lotus Scrotes could,
> because the document never really leaves the database, but how would you
> build a system that reliably worked in a heterogenous environment like a
> small-medium office, that actually worked, and you could sell to people and
> still retain your soul?
>
> --

I tend to think that such devices are probably more "security theatre"
as Bruce said it in his keynote, as it is hard to do reliably. If you
allow users adhoc access to mail or web browsers, while you can catch
sequences of numbers like 1234 if you are watching for credit card
numbers, are you watching for one two three four, onetwothreefour,
eentweedrievier and I,II,III,IV as well? This is simple encryption
that people can easily detect, but with modest obfuscation are
possibly hard for automated systems to correctly detect. In order to
effectively limit data leakage I think you need :-

1. Limit user access to the data on a "need to know basis". For
instance,credit card numbers are probably best stored securely once,
and never presented back to users (with transactions being done with
no direct user intervention). As long as administrative access to the
data is very limited (or even nigh on impossible is using end-to-end
hardware encryptors) then the possibilities for stuff getting out are
limited. In fact, in this case the DLP systems mentioned become
useful, in that leakage is likely to be only the result of inadvertant
configuration errors, rather than adhoc and wanton transmission. When
system users do need read access to sensitive data then it is useful
to have unreputable auditing, so that people know they are being
watched. (I understand this approach in police databases is now
employed and has reduced inside abuse of data access substantially -
as least that is what they tell us :-) ) You often don't need all the
data on a customer to be presented to a customer service consultant,
only that for the particular part of the transaction at hand.
2. Have totally separate systems that handle very private and
sensitive data. Ideally this goes all the way to networks and
workstations. If you do need to save having totally redundant systems
you could employ techniques like data diode
(http://en.wikipedia.org/wiki/Unidirectional_network) or Nettop
(http://en.wikipedia.org/wiki/NetTop - yes, my employer sells a
product based on the latter ). None of these prevent the "pencil and
paper" leakage of data, but at least they limit the wholesale transfer
of data between sensitive and public environments. If the data is
contained behind the glass screen only they you are much more likely
to keep tabs on it. (In fact one of the premises behind the idea of
"blade PCs" with very thin desktop clients is trying to keep the data
in the data centre and not on the desktop (apart from on the screen).

That being said, a reverse-snort like appliance sounds like a cool idea!

Regards, Martin
-- 
SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/
Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html

Reply via email to