See also: IRC log
<scribe> Scribe: Jeff Waters
<scribe> ScribeNick:jeffw
jeffw: We will have guests today
to help us learn about security ontologies from 3 sigma
research
... Welcome to all.
... Jim Dike, Aaron Wheeler and Michael Winburn are joining us
from 3 sigma research
michael: I'll probably turn it over to jim dike for some background
jimdike: we've tended to focus on
ontologies related to security domain, how to model and
represent entities that are significant in that area
... what entities are relevant for security and dissemination
of information, there are analogies to government work and all
of the constraints and restrictions in that environment, but
also in the commercial world
... so we settled on some simple concepts of defining a person
and characteristics that are needed for defining a person with
respect to handling info that needs to be secure
... What I mean by "secure" is bound by what types of access
that person needs to have and what kind of characteristics
describe the groups that define sets of restrictions, then they
can be applied to the info
... but we've also applied them to transport media. Initially,
it was thought they might be limited to specialized and private
networks, but with the explosion of vpn, the concept of what
info should go out on what vpn complicates
... people, media and transport mechanisms, how to represent
all of that consistently and how to determine if something
really should go from A to B
... it's a simple model but can get complicated as you scale
it.
... in a commercial setting, these might represent business
relationships, departments, for example a finance group, a
contracts group, an engineering group a research development
group that are only interested in some info and not
others
... groups of persons, groups of data items and groups of
transport mechanisms, so we ended up settling on and
researching how to apply ontologies and reasoners to that
... reasoners to determine consistency and validity of those
relationships, and to identify the case where we would want to
transport and where there are inconsistencies and
contradictions
... that's where we really start to see some interesting
things. It's the things you don't immediately see that are
interesting. Using ontologies and reasoners, we use OWL and OWL
DL to represent the ontology.
... Then with reasoners, you can see those relationships not
immediately evident. We're itnerested in ensuring some info
doesn't go where it shouldn't go.
jeffw: it seems like this is a nice generic model which can apply not just to traditional classification, but also for things like passing information from one city to another that borders on a county or state with different policies
jimdike: yes, often you have
aggregations of data items, some might be more public, some
might be more sensitive, but when you begin to aggregate pieces
of info from multiple sources, so in business world,
... say you have a coaltiion of businesses working together
with different disclosure agreements, so say two are more open
so the info is "public" but for another entity it might be more
closed
... how do you disseminate different portions of data items,
and can the aggregate go to everyone or only pieces of the
aggregate go to everyone, from a more simple to a more complex
or complex to individual, ifyou model it properly, you can go
in both directions
... you might have an aggregate of business info that wants to
go out, as a whole, you might say too sensitive, so it doesn't
go out, but then again internally organizations don't think of
how to divide it up and go to different partners
... so we've looked at that in our models.
jeffw: what reasoner do you use?
jimdike: we try against multiple
reasoners
... we treat them like black boxes, there are slight
differences and we'll test against multiple reasoners, so we
want it to be robust, we constrain the testbed to pellet,
fact++, we use protege and they have a new one
... there also is an engine JENA that has a reasoner that they
have. We tend to restrict ourselves to OWL DL versus OWL Full
to ensure the models are deterministic.
... open world v. closed world, but for some things you want
them to be deterministic
michaelWinburn: we can internally talk about what we can share
jimdike: it's fair to say that
it's part of our current work to define that line between what
can be public and what is proprietary, we aren't the only ones
who can define a security ontology
... we want to define that line, then people can drop into it,
the ontology itself is an open data standard, then reasoning
and things behind it would be internal or proprietary, but to
represent decisions in a way that would allow individual
vendors to solve interesting problems, we could think about
that
... we could think of a representation that is good for
semantic uses
jeffw: it seems like a public standard and then proprietary uses would be great, I think that might be a business model, you are certainly welcome to join w3c and participate or participate as an invited expert
jimdike: yes we would be interested
jeffw: Would Don, Eva, Piotr like to ask a few questions?
Don: yes, I appreciate you presenting this in a public forum and these folks have some expertise in ontologies, and we can follow up in private as well
eblomqvi: Yes, thanks, even if some info is public when you aggregate different types of information, the orginal data is partially public, partially not, these are important problems and interesting to hear people are working on this
jimdike: We've been working in this area for 5 years now?
pitor_nowara: this could be a good use case in my opinion
jeffw: it seems like the mathematical foundation is a strong support for the reasoning that is a selling factor for your approach, is that correct?
jimdike: I mentioned before that
we made a decision to stay with OWL DL, I have been to at least
one conference where there was an academic discussion about how
OWL DL is insufficient to represent the cases where real
decisions can be made
... what you are doing they say is throwing the baby out with
the bathwater, but we found out that there is a medium ground,
you can define models and use reasoners to a certain point and
then you add processing on top, you can get some desired
results
... the model is good for the purpose it is built for and that
is good enough, to say that it has to handle the complete
domain is not a good approach. That's an important lesson we
learned.
<Don> Jeff - Need to run to another meeting
<Don> thanks for putting this together
<Don> and fostering a great discussion
jimdike Nick Drummond with Univ of Manchester in UK said this stuff is hard, ontology modeling is not for the faint of heart, keep in mind the intent and constrain yourself to that.
jimdike: for security people, the
concern is is it deterministic, how do you prove this is true
in the complex case and that takes a little more time, we're
still in process of how to communicate that to a lay
person
... it can add assistance to their process of verification, it
can help speed up the work of that person
jeffw: do you have anything you can point to as a reference?
jimdike: we can get back with you on that.
jeffw: any update on paper or modeling problems?
eblomqvi: just a week and a half away, we present it as a poster, I did send it to the authors, and we need to check the format and any inputs you have. I wanted to ask if we can have some graphics and I was thinking about the prototype system, is there anything more we can show on the poster, some diagram or structure or something, that would be nice to have
jeffw: we don't have any diagram for the prototype right?
eblomqvi: any flow diagram or user interface diagram
jeffw: Piotr, any update on criteria model?
piotr_nowara: I was doing more
examples, but I'll let you know when I finish that stage of my
testing, I would like to share my model, perhaps on
ontologydesignpatterns.org website.
... I don't know if it's good enough
eblomqvi: yes,
ontologydesignpatterns would be a good place and don't worry
about if it's not the final version, idea of the portal is
something that people can contribute to and comment on, it's
like wikipedia, you can add to it even if not sure
... we hope to increase discussion activity on the portal and
at least people can see it and people can contact you directly,
so please publish it on the portal
jeffw: We're out of time for
today. I'd like to thank Michael, Jim and Aaron for
participating today and educating us on their approach to
developing an ontology to model security for information
exchange. We will definitely follow-up.
... Also thanks to Eva, Piotr and Don for calling in. Our next
meeting will be in two weeks and you are all welcome to
participate. We'll look forward to developing our decision
model and applying it to our various use cases. Thanks.
This is scribe.perl Revision: 1.135 of Date: 2009/03/02 03:52:20 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Found Scribe: Jeff Waters Found ScribeNick: jeffw WARNING: No "Topic:" lines found. WARNING: No "Present: ... " found! Possibly Present: Don ScribeNick eblomqvi jeffw jimdike michael michaelWinburn piotr_nowara pitor_nowara You can indicate people for the Present list like this: <dbooth> Present: dbooth jonathan mary <dbooth> Present+ amy Agenda: http://www.w3.org/2005/Incubator/decision/wiki/Decision_Mtg_16_Agenda Got date from IRC log name: 28 Oct 2010 Guessing minutes URL: http://www.w3.org/2010/10/28-decision-xg-minutes.html People with action items: WARNING: Input appears to use implicit continuation lines. You may need the "-implicitContinuations" option. WARNING: No "Topic: ..." lines found! Resulting HTML may have an empty (invalid) <ol>...</ol>. Explanation: "Topic: ..." lines are used to indicate the start of new discussion topics or agenda items, such as: <dbooth> Topic: Review of Amy's report[End of scribe.perl diagnostic output]