Re: [logs] firewall logging and rulesets

From: Brian Ford (brford@private)
Date: Mon Oct 27 2003 - 10:36:25 PST

  • Next message: Raffael Marty: "Re: [logs] firewall logging and rulesets"

    Raffy,
    
    As it is Syslog the PIX has the traditional severity levels.  I've found 
    that 0-7 is better reduced to three bins.  Again, this is my own personal 
    interpretation.  I don't read tea leaves but i do look at Firewall logs.
    
    I think that most device vendors have been trying to respond to customers 
    requests for instrumentation in the log.  So we have devices that say "I 
    see an X attack".  Great, we've hard coded known attacks!  What we need is 
    to have the tools so that we can put together lower level reports or 
    primitives from a variety of devices.
    
    Clearly this isn't going to be for everyone (yet).  But recall that covered 
    wagons had -no- gauges.  The early automobiles (1915) had two gauges; a 
    speedometer and a tachometer.  By the 1950s they have five.  The computer 
    in my 2003 auto checks over 100 different readings (but it doesn't show me 
    100 gauges).  Right now I think we are at the point where we are just 
    breaking away from wanting to look at all the gauges.
    
    Liberty for All,
    
    Brian
    
    At 09:57 AM 10/27/2003 -0800, Raffael Marty wrote:
    ><Disclaimer: I work for a security information management (SIM) product
    >company/>
    >
    >Brian,
    >
    >Thanks for you extensive answers!
    >
    > > >Does the PIX actually indicate these buckets (or similar ones) in the 
    > events?
    > >
    > > No
    > >
    > > >Could I use the event-id ranges for that?
    > >
    > > I do this via message ID.  I am in the process of writing this up.
    >
    >Why isn't there a field in the message indicating the bucket? Something
    >like an event-category or a severity!
    >
    > > >Or is that a feature which will be added in a future release?
    > >
    > > Doubtful.  We're trying to put the features in there so different designs
    > > and reporting requirements will be supported.
    >
    >I think I can't quite follow. Logs are there to hold information about
    >the status of a device and actions it took. Why isn't it a reporting
    >requirement to have those buckets?
    >
    > > > > - I like to count the normal operations messages and make sure that I
    > > > > create some baseline (between x-y per hour or y-z per day).  This 
    > baseline
    > > > > is a kind of load gauge for the device.
    > > >
    > > >Good point. Now you just need the tool to do this.
    > >
    > > Ding, ding, ding! Winner.  Absolutely.  I think we need to ask a bunch of
    > > tool vendors to support this.
    >
    ><Vendor/Product-Plug>
    >Among other features, that's what we (ArcSight) do in our product. You
    >can have so-called datamonitors which observe the amout of data coming
    >from a certain device. If the amount of events starts to increase or
    >decrease too much, you can be notified or you can take any other action
    >you define.
    ></Vendor/Product-Plug>
    >
    > > > > - The "irregular" messages usually require correlation with other
    > > > > conditions or events.  If someone were auditing what I was doing they
    > > > would
    > > > > ask about those.  I should have an explanation.  How often I do that
    > > > > depends on what I can get away with (audit interval).
    > > >
    > > >For this you normally need some more context information; basically the
    > > >capability to correlate with other event sources!
    > >
    > > Yes.  And that is tremendously frustrating since no one seems to do much
    > > the same way.  But it is do-able.
    >
    ><Second Vendor/Product-Plug>
    >Well, we are! I think this is where everyone will go in the future! You
    >can only understand your network and know what is going on in real-time,
    >if you use _all_ your sources of information to help you draw the
    >picture!
    ></Second Vendor/Product-Plug>
    >
    > > >I can try to be the "someone to come up with a [really good?] list":
    > > >Tell me everything the firewall does! That's all I want! Then assign a
    > > >severity to the events and let me decide which ones I wanna look at.
    > > >Only if I see all the events, I can draw myself an adequate picture of
    > > >what is happening across my network.
    > >
    > > I think that this is great.  I suggest you need to use the Firewall in the
    > > network you operate and others you design. Alot of the work here isn't 
    > just
    > > listing all the features.  It's how the features are used together to 
    > solve
    > > various design issues.
    >
    >What I wanted to emphasize is that most of the device-vendors just log
    >if a packet was dropped or if it was passed. That's just not enough
    >information. I would like to have much more information. If you filter
    >IP fragments and the packet is dropped as a result of that filter, I
    >would like to see that indicated in the logged event. Furhtermore, I
    >need to know if someone tries to access the firewall, when it fails
    >over, etc. Therefore I really appreciate the extent at which the PIX
    >logs.
    >
    >Thanks
    >
    >         Raffy
    >
    >--
    >
    >Raffael Marty, CISSP                          raffael.marty@private
    >Security Engineer                           Content Team @ ArcSight Inc.
    >1309 South Mary Ave.         Sunnyvale, CA 94087          (408) 328 5562
    >
    >Raffy's opinions are not necessarily ArcSight's policy.
    
    
    Brian Ford
    Consulting Engineer, Security & Integrity Specialist
    Office of Strategic Technology Planning
    Cisco Systems Inc.
    http://www.cisco.com/go/safe/
    
    The opinions expressed in this message are those of the author and not 
    necessarily those of Cisco Systems, Inc.
    
    
    
    
    _______________________________________________
    LogAnalysis mailing list
    LogAnalysis@private
    http://lists.shmoo.com/mailman/listinfo/loganalysis
    



    This archive was generated by hypermail 2b30 : Mon Oct 27 2003 - 10:44:41 PST