Re: [logs] Problems/Questions on PEO

From: Devin Kowatch (devinkat_private)
Date: Mon Apr 01 2002 - 12:26:51 PST

  • Next message: Sweth Chandramouli: "Re: [logs] Problems/Questions on PEO"

    On Sun, Mar 31, 2002 at 05:16:06PM +0200, Shane Kerr wrote:
    > On 2002-03-29 12:24:16 -0800, Devin Kowatch wrote:
    > > 
    > > The second problem is that it is not possible to determin where the
    > > log was altered.  If an attacker inserts or alters a message at time t
    > > then all messages after time t will not be verifiable.  But because
    > > the intermediate hash values are not saved it will also be impossible
    > > to verify messages _before_ time t.  
    > 
    > My understanding is that we save the original key, meaning that the
    > entire chain of hash values and keys can be generated, given the log
    > entries.  Using this information you can verify all logs made before the
    > compromise are unaltered.
    
    This is true, but with a bogus message (i.e. not part of the hash chain)
    anywhere in the log file, it is impossible (or at least difficult) to
    figure out which message is bad.  All you have is a starting key, and
    the final hash value.  So if we have a starting key, K, generated at 
    time t=0, and the last hash value, Kt, generated at time t.  With these
    we can tell if a message was inserted, altered, or deleted, but not
    which message.  
    
    When it comes to post compramise verification, the compramise is likely
    to have occured at time s, where 0<s<t.  So it would be possible to
    figure out that the attacker changed a log entry, but not which one or
    when.  With a some careful deletions the records of the attack can be
    removed, and new records of a fake attack (from another host) can be
    inserted.  The fake records would get inserted to throw the
    investigation off, because it's known that someone altered the logs.
    
    That is assuming of course that there is no central log server.  This 
    is a, hopefully, invalid assumption.
    
    Does that make a little more sense?
    
    thanks,
    -- 
    Devin Kowatch
    devinkat_private
    
    ---------------------------------------------------------------------
    To unsubscribe, e-mail: loganalysis-unsubscribeat_private
    For additional commands, e-mail: loganalysis-helpat_private
    



    This archive was generated by hypermail 2b30 : Mon Apr 01 2002 - 14:11:37 PST