IDS mailing list archives

Re: Question on resources needed to manage IDSes


From: Terence Runge <terencerunge () sbcglobal net>
Date: Tue, 2 Dec 2003 11:12:18 -0800 (PST)

My two very unscientific cents on the subject.

Analyst to employee ratio 1:3000
Analyst to sensor ratio 1:30
Analyst to bad guy ratio 1:1 ymmv

Those are some loose numbers that have stayed fairly
consistent in my experience as an analyst with both
public and private sectors.

Some of the key missing factors here include
analyst:server, analyst:class, and
analyst:environment. As an example, more time is spent
"researching" the anomolous activities of 7000+
employees running windows than the 100,000+ servers
running *nix.

Other factors are attrition and culture. This is more
along the lines of behavior analysis, but it keeps an
analyst busy. Most new employees who are going to
involve themselves in some sort of nefarious activity
will, in my experience, typically do so over a three
month time sequence. Again, the bad guy ratio, which
may be 1:1 but can take months of your time.

Three things which will cut down on the amount of time
an analyst spends staring at the ids are clear policy,
well written rules, and knowledge.

Terence



-----Original Message-----
From: Jack Whitsitt (jofny)
[mailto:seclists () violating us]
Sent: Monday, December 01, 2003 7:14 PM
To: kgeorgiades () toplayer com
Cc: focus-ids () securityfocus com
Subject: Re: Question on resources needed to manage
IDSes


The idea that IDS's are black boxes which can be
dropped in...and that there
is some rule of
thumb for their behavior...is one of the biggest
driving factors in
generating huge volumes of
logs and false alarms.

An IDS's job is to report when a defined security
policy has been violated.
How well has that
policy been defined? How well have the sensors been
configured to be aware
of the policy and
detect violations?

The number of resources which need to be dedicated
on a per-sensor basis is
a direct function
of the security policy that you've created ahead of
time and how much effort
you've put in to
mapping your tools to it.

In addition to the background information, you need
to keep in mind that not
all IDS's are
created equal. One choice is *not* as good as
another.  If my tool breaks
down or doesn't
provide me with information that I need, it's going
to negatively impact my
turnaround time.

There might be guidelines which could be developed
regarding analyst-sensor
ratios, but not
without a lot of surrounding context.  The tools and
techniques are not yet
standardized and
operationalized well enough to provide easy (if any)
statistics.  There is
no evil bit that
gets set in network traffic...

I know you were probably looking for even generally
applicable stats, but
IDS management isn't
as far along in it's business lifecycle as, say, a
technical support call
center or a desktop
support environment...(IMO)


-Jack Whitsitt


Everyone seems to be talking about the large
volume of alarms and logs
produced by IDSes.
Managing IDSes and dealing with false alarms seems
to be an issue that all
IDS vendor are
trying to address.

Has any one of you seen any data on how many
analysts (resources) are
needed to manage IDSes
in enterprises?

I am looking for a rule of thumb, something like
this:
1-5 IDS sensors - 1 Analyst
5-15 IDS sensors -2 Analysts
15-50 IDS sensors- 3 Analysts
1 Analyst for every 30 additional IDS sensors.

I will appreciate any feedback that I can get.

Thanks,

Kyriacos (Ken) Georgiades
Senior Director, Product Line Management
Top Layer Networks, Inc
Tel: 508 870 1300 x 231
Cell: 508 783 5988
Fax: 508 870 9797
Email: kgeorgiades () toplayer com
www.toplayer.com




---------------------------------------------------------------------------


---------------------------------------------------------------------------





---------------------------------------------------------------------------
---------------------------------------------------------------------------

---------------------------------------------------------------------------
---------------------------------------------------------------------------


Current thread: