RE: Views and Correlation in Intrusion Detection

[David Markle <davidmarkle () comcast net>]

Well, since we are speaking in idealistic terms, retrofitting a system of
this magnitude in a truly global enterprise would be very difficult and the
scaling issues would definitely come into play.  My thoughts on this are
that the global model must have a hierarchical tiered approach where the
event aggregation (central collection PLUS event
linking/counting/suppression) and correlation takes place further down the
tree prior to it boiling up to the top level monitor.  This is the only way
to distribute processing and deal with scalability.  So you're absolutely
correct, no one system will handle the load (i.e. correlation for the
mainframe).

I agree with your vendor standardization comments.  They are generally NOT
willing to spend the development $$ on something that does not produce
revenue first (no offense vendors, but its a revenue based world ;)  ).
Therefore, as we are seeing with the several products out there (Arc-site,
etc....), log agent listeners are developed just for this "vendor" specific
purpose (aggregation and normalization).

There are a whole lot of smart people out there and the problems can be
resolved.  The scalability issue can be resolved via the hierarchical tiered
approach, add levels of duplicate alert suppression, bandwidth throttling,
and queuing and the issue is pretty much resolved.  (remember ...we're being
idealistic here ...)

I think the greatest growth potential in the near future is the intelligence
behind the correlation.  Aggregating and normalizing the logs is not a huge
problem.  Companies with $$ with be able to hire the mathematicians, and
statisticians to create the needed logic behind the concept.  Hopefully, our
Open Source community has a few of those types out there to help push the
freeware movement.

as always, my $.02

David Markle


-----Original Message-----
From: Sekurity Wizard [mailto:s.wizard () boundariez com]
Sent: Wednesday, June 25, 2003 11:03 PM
To: DAVID MARKLE; Blake Matheny
Cc: focus-ids () securityfocus com
Subject: RE: Views and Correlation in Intrusion Detection


David,
        Your are all absolutely correct - correlation is the gold
medal...right now everyone in the industry is praying for bronze at
best.  The one glimmer of hope I see are products out there, and I don't
remember the company name right now, that aggregate hundreds of gigabits
of logs per hour and try to make sense of it all.  The question them
becomes one of scalability...assuming we take for granted someone CAN
write an engine that processes this sort of data in a sane manner.
        Scalability, in the form of the type of environment I work at is
insanely large.  We have umpteen numbers of DS3's, countless T1's and
thousands of pipes to and from segments we aren't even *aware of*...not
to count the couple of hundred (close to 1,000) firewalls that are out
there.  Now, let's say we put a couple of these boxes (~50Mbit/sec each)
to the test in my environment.  There STILL NEEDS TO BE A CENTRAL
PROCESSOR...otherwise, we're left with the distributed view - which we
don't want, right?  Is it realistic to think there is such a scalable
system that can process hundreds of gigabits of data per second,
aggregate it all, normalize it, and correlate too?  I dare say not at
this point...unless we come up with some sort of standard, "XML for
security devices" that makes the processing and data crunching
easier....but the problem there is I don't see Checkpoing, Cisco,
Enterasys, and ISS (and others) getting together on this any time
soon....

        So scalability is our main opponent as I see it...because at the
end of the day - the only attack that counts is the 1 in 100,000,000
that sent that single UDP packet that triggered a shutdown of the entire
network due to SQL Server port floods...right?

Sleep well... :)

-----Original Message-----
From: DAVID MARKLE [mailto:davidmarkle () comcast net]
Sent: Tuesday, June 17, 2003 1:49 PM
To: Blake Matheny
Cc: focus-ids () securityfocus com; davidmarkle () comcast net
Subject: Re: Views and Correlation in Intrusion Detection


Blake, I agree with your sentiments regarding correlation and have more
to add.

The point of correlation is the value it adds to mostly autonomous,
unreviewed, and meaningless data. (The folks that disagree with this
line must have economically independent budgets with staffing
consisting of superstar (I applaud you)).  Who reviews the firewall
logs?  I don't.  We have over 500 global firewalls.  The point here is
(as you stated) AUTOMATION.  But it does not stop there.  That data has
to be normalized and applied towards something.  The correlation piece
adds that middleware "something".  An IDS alert is ONLY relevant if the
firewall permits the traffic through.  To further the comment, and
attack signature tripped for (known attack) xyz, is ONLY relevant when
the attacked host is vulnerable to xyz.  This is the ultimate job of
correlation.  If the above surrounding conditions are true, the
severity of the attack becomes increased to critical, otherwise it is
informational only.  There are also netops statistics that should be
considered security related (and monitored).  Baseline your bandwidth,
averaged over 12 months.  Normal increases in business offerings are
roughly 5 percent per month.  Since there was no change control this
past weekend (to relate), why did you see a spike in bandwidth by 17
percent ????  Why is tcp 2148 increasing on your global perimeter over
the past 3 days? These are relevant questions.  Without the collection
and aggregation of the appropriate data, we run the operations in the
dark.

With regards to the state of correlation, I still think its an infancy
issue.  Historically, I believe that the industry (tech folks) has been
extremely focused on growth development and deployment of the
technology (firewalls, IDS-(H/N), etc.).  Firewalls have been around
for awhile and have matured to a point of plateau (mostly).  IDS is now
in "the growth phase" (with heuristic, anomaly, signature, blah, blah,
blah), and all that hype.  I really think that the industry had
recently realizes that we are now overwhelmed with too much data. Now
everyone is scrambling to catch up .....

David Markle
davidmarkle () comcast net
davidmarkle () elephantfoot org




----- Original Message -----
From: Blake Matheny <bmatheny () mkfifo net>
Date: Tuesday, June 17, 2003 1:32 pm
Subject: Views and Correlation in Intrusion Detection

> Two areas that I have recently been doing research in, are views
> and their
> connection to correlation techniques. In terms of systems, given
> some event,
> the information we get about the occurrence of such an event comes
> to us in
> the form of either a primary or a secondary view. Information
> about secondary
> views typically come to us from applications such as firewalls and
> ID systems.
> Primary information usually is received from the application actually
> processing this data for use. For instance, an ID sensor may
> produce an alert
> about some traffic. However, this is a secondary view of the event
> and needs
> to be correlated with other, relevant information. So of course
> firewall logs
> might be checked, to see if traffic actually passed that
> corresponds to the
> event in question. This is also a secondary view, so a third place
> is checked,
> the applications logs.
> There are really several issues here. First of all, a tremendous
> amount of
> time is being spent, trying to correlate all the relevant
> information. This is
> something that _can_ be automated. Second, the applications logs
> may not be
> trustworthy. Third, and to me, most importantly, is the fact that
> this is such
> a 'basic' thing that people using ID systems have to do, and there
> is no piece
> of software yet that does this.
> So something we have been working on, is a system to deal with
> this basic
> type of scenario. This will entail data transformations into an
> intermediarylanguage, an event description language, offline state
> analysis and several
> other components (there is more information at
> http://www.nongnu.org/babe/).If you spend some time thinking about
> everything involved to do this in a
> scalable fashion, it's an enormous task (I said basic, not
> simple). What I am
> finding frustrating, is that much of the base research has not yet
> even been
> done. Much of the research that has been done, is either too
> primitive or too
> impractical to be implemented. Is this due to the infancy and
> immaturity of
> the field, do people not see this as being feasible and therefor
> aren'tspending the research time, or is this simply too far down
> the line? In any
> case, feedback welcome. Thanks.
>
> Cheers,
>
> -Blake
>
> --
> Blake Matheny           "... one of the main causes of the fall of the
> bmatheny () mkfifo net      Roman Empire was that, lacking zero, they had
> http://www.mkfifo.net    no way to indicate successful termination of
> http://ovmj.org/GNUnet/  their C programs." --Robert Firth
>
> -------------------------------------------------------------------
> ------------
> Attend the Black Hat Briefings & Training, July 28 - 31 in Las
> Vegas, the
> world's premier technical IT security event! 10 tracks, 15
> training sessions,
> 1,800 delegates from 30 nations including all of the top experts,
> from CSO's to
> "underground" security specialists.  See for yourself what the
> buzz is about!
> Early-bird registration ends July 3.  This event will sell out.
> www.blackhat.com---------------------------------------------------
> ----------------------------


------------------------------------------------------------------------
-------
Attend the Black Hat Briefings & Training, July 28 - 31 in Las Vegas,
the
world's premier technical IT security event! 10 tracks, 15 training
sessions,
1,800 delegates from 30 nations including all of the top experts, from
CSO's to
"underground" security specialists.  See for yourself what the buzz is
about!
Early-bird registration ends July 3.  This event will sell out.
www.blackhat.com
------------------------------------------------------------------------
-------


-------------------------------------------------------------------------------
Attend the Black Hat Briefings & Training, July 28 - 31 in Las Vegas, the 
world's premier technical IT security event! 10 tracks, 15 training sessions, 
1,800 delegates from 30 nations including all of the top experts, from CSO's to 
"underground" security specialists.  See for yourself what the buzz is about!  
Early-bird registration ends July 3.  This event will sell out. www.blackhat.com
-------------------------------------------------------------------------------