Re: [logs] log review policies

From: Matthew G. Marsh (mgmat_private)
Date: Sat Oct 20 2001 - 08:13:00 PDT

  • Next message: Chris Lonvick: "Re: [logs] any syslog implementations of draft-ietf-syslog-reliable?"

    On Fri, 19 Oct 2001 peff-loganalat_private wrote:
    
    > On Fri, 19 Oct 2001, Matthew G. Marsh wrote:
    >
    > > structure methodology. To that end I am also a big promoter/user of SNMP.
    > > We make extensive use of SNMPv3 in many of our managed security
    > > structures. This is the framework behind the following technique:
    >
    > I haven't used SNMP in years...what defenses does v3 provide against
    > spoofing?  Assuming that strong cryptographic authentication could be
    > used, then this sounds like a great idea for doing central system
    > management.
    
    Authentication via SHA or MD5 plus a separate DES encrypt of the data
    packet using a (optional) different passphrase. Essentially:
    
    SNMPv1 - clear text community
    SNMPv2 - ability to lock down somewhat to users/groups
    	 but still clear text
    SNMPv3 - User/Group lockdown of v2 PLUS:
    	- MD5/SHA hash of user authority passphrase
    	- DES encrypt of data + optional different passphrase
    	- INFORM trap - which is a trap that must be replied to
    		before the client gives up sending.
    	- TCP capabilities
    
    I actually use NET-SNMP for all my research and usage as that gives me
    several additional security measures. Such as:
    
    - Daemon process on "client" (managed station) only needs the
    	SHA & DES passphrases specified ONCE on initial configuration
    	startup. It then remembers these values in a hash that is tied
    	to the machine it was generated on.
    
    This is true of both snmpd and snmptrapd - this is very cool as someone
    cannot use a violated client machine to crack your servers or other
    clients.
    
    - TCP_WRAPPERS support for selective reception
    
    - TCP connections and INFORM trapping provides additional reliability
    
    standard disclaim + -> http://www.net-snmp.org
    
    > I suppose you could also just do cryptographic signing at the layer
    > above SNMP. I assume you're also working on some sort of timeout at the
    > central server...that is, if you don't receive a hash in N seconds, an
    > alarm is tripped.
    
    Correct - timeout is standard network management procedure. Usually
    manager stations in SNMP have both regular interval checking (polling) of
    clients and reception capability for traps. The check I use are part of
    the standard polling structure. And I actually do much deeper structures
    on the checks with some randomness thrown in to reduce the average window.
    
    > > You have to crack one of these systems within 5 minutes in such a manner
    > > as to change the OOB logging, AND disable the SNMP trapping mechanism, AND
    > > disable the host IDS mechanism, AND finally make sure that you send back
    > > appropriately spoofed hashes.
    >
    > This would be a perfect scenario for a ``hacking'' scene from a
    > Hollywood film. :)
    
    Yes - I think with the appropriate skill set and a very well informed
    attacker it would be feasable - but then technically speaking anything is
    feasable in the security world. The hard part is getting across the
    concept of acceptable risk. ;-}
    
    > > Nothing is perfect but experience teaches that several short barbed wire
    > > fences separated by moats is much much better than one large fence...
    >
    > Agreed.
    >
    > -Jeff
    >
    >
    > ---------------------------------------------------------------------
    > To unsubscribe, e-mail: loganalysis-unsubscribeat_private
    > For additional commands, e-mail: loganalysis-helpat_private
    >
    
    --------------------------------------------------
    Matthew G. Marsh,  President
    Paktronix Systems LLC
    1506 North 59th Street
    Omaha  NE  68104
    Phone: (402) 932-7250 x101
    Email: mgmat_private
    WWW:  http://www.paktronix.com
    --------------------------------------------------
    
    
    ---------------------------------------------------------------------
    To unsubscribe, e-mail: loganalysis-unsubscribeat_private
    For additional commands, e-mail: loganalysis-helpat_private
    



    This archive was generated by hypermail 2b30 : Sat Oct 20 2001 - 09:29:13 PDT