Re: Intrusion Detection

From: Aleph One (aleph1at_private)
Date: Wed Apr 15 1998 - 17:46:04 PDT

  • Next message: Sheila Or Bob (depends on who is writing0: "Re: When to do something about detected attacks (was Re: how to do...)"

    On Wed, 15 Apr 1998, Marcus J. Ranum wrote:
    
    > I dunno how many of the folks on this list remember Fred Cohen's
    > "intrusion detection" system that he used to run on all.net. If
    > you tried to Telnet to his system, it would look up the registered
    > contact for your domain and E-mail them a nastygram that someone
    > had just tried to break in to his system from your workstation.
    
    I certainly do. I also remember how people created web pages with an
    embedded telnet:all.net link for people to stumble on. Everyone (except
    Fred maybe) got a kick out of it.
    
    > I'm starting to convince myself that I want to implement IDS
    > as policy-based traps (a la Raiders of the Lost Ark -- if someone
    > runs teardrop on me I want a big rock to fall on them) backed
    > with passive sensors (microwave/PIR packet suckers) to catch
    > anything that sneaks past. There are so many physical security
    > analogies for how to do this right -- it's all beginning to come
    > clear for me now.
    
    This is in essence and IDS that is both a static ADS and MDS.
    The ADS part looks for network activity that does not match its profile of
    the network (your policy). Its static because it does not learn this
    policy from the network but from the configuration interface. The MDS part
    is what looks for attack signatures. This may be part both of the traffic
    that matches your network profile and that which does not. 
    
    The MDS should attempt to determine if the attack is has just detected has
    been successful or not. This is simple to know in certain attacks (DoS,
    buffer overflows, etc) and more difficult in others. The IDS would
    classify events into either warning or alarms. Any breach of your network
    profile or an attack signature match that was or may have been successful
    would be reported as an alarm. Unsuccessful attacks would be warnings.
    
    > mjr.
    > --
    > Marcus J. Ranum, CEO, Network Flight Recorder, Inc.
    > work - http://www.nfr.net
    > home - http://www.clark.net/pub/mjr
    > 
    
    Aleph One / aleph1at_private
    http://underground.org/
    KeyID 1024/948FD6B5 
    Fingerprint EE C9 E8 AA CB AF 09 61  8C 39 EA 47 A8 6A B8 01 
    



    This archive was generated by hypermail 2b30 : Fri Apr 13 2001 - 12:54:58 PDT