On 9/25/13 12:00 AM, Arlo Barnes wrote:
It seems unacceptable that a statement could stand like "X is 52%
classified" where X is not an aggregate disconnected set of
things, but some single fact in context. It would be just muddy
guidance. The fact in context can be disclosed to a specific
audience, or it cannot. It can't be disclosed 52% of the time.
Often, though, there is confusion about what the parameter to be
discretized is. For example, you might use 'facts' as the parameter,
and say something like "52% of the facts about Project X are disclosed
in the press release."
Right. That's why I qualified it with "aggregate disconnect set of
things". I mean that the hypothetical fact or relation was a primitive
not a composite, e.g. "bin Laden believed to be in Abbottabad" (prior to
his death). Another 99 facts, of which, say, 51 of them were, say,
about locations of Galeb-4 fighter aircraft in the former Yugoslavia in
1999 prior to NATO bombings (info now available at www.foia.cia.gov, but
then secret), and another 48 which were about the popular flavors of ice
cream in Oklahoma City.
In the case of one person probing sensitive personal information of
another person, the latter might say "I'm not comfortable talking about
that" or modify/truncate the details of story on the fly to not reveal
their discomfort nor their information.
In a triple store database, a query for relations would return different
rows depending on who was asking, and no triples could be added for a
lower security level if they were derived from queries made at a more
restrictive level. Probably simply limiting records isn't sufficient --
a triple store front end might also sometimes need to invent proxy
information (cover stories) to maintain self-consistency.
Quantitative information is tricky, since anything that is revealed is a
stake in the ground for future queries, e.g. Glen's example of the black
budget could be deduced within wide error bars by starting with the
country's GDP as an upper bound. Bit it is absurd to make the GDP a
secret.
Marcus
============================================================
FRIAM Applied Complexity Group listserv
Meets Fridays 9a-11:30 at cafe at St. John's College
to unsubscribe http://redfish.com/mailman/listinfo/friam_redfish.com
