RFC 1858 (rfc1858) - Page 2 of 10

Security Considerations for IP Fragment Filtering

RFC 1858    Security Considerations - IP Fragment Filtering October 1995


2. Filtering IP Fragments

   IP packet filters on routers are designed with a user interface that
   hides packet fragmentation from the administrator; conceptually, an
   IP filter is applied to each IP packet as a complete entity.

   One approach to fragment filtering, described by Mogul [1], involves
   keeping track of the results of applying filter rules to the first
   fragment (FO==0) and applying them to subsequent fragments of the
   same packet.  The filtering module would maintain a list of packets
   indexed by the source address, destination address, protocol, and IP
   ID.  When the initial (FO==0) fragment is seen, if the MF bit is set,
   a list item would be allocated to hold the result of filter access
   checks.  When packets with a non-zero FO come in, look up the list
   element with a matching SA/DA/PROT/ID and apply the stored result
   (pass or block).  When a fragment with a zero MF bit is seen, free
   the list element.

   Although this method (or some refinement of it) might successfully
   remove any trace of the offending whole packet, it has some
   difficulties.  Fragments that arrive out of order, possibly because
   they traveled over different paths, violate one of the design
   assumptions, and undesired fragments can leak through as a result.
   Furthermore, if the filtering router lies on one of several parallel
   paths, the filtering module will not see every fragment and cannot
   guarantee complete fragment filtering in the case of packets that
   should be dropped.


   Fortunately, we do not need to remove all fragments of an offending
   packet.  Since "interesting" packet information is contained in the
   headers at the beginning, filters are generally applied only to the
   first fragment.  Non-first fragments are passed without filtering,
   because it will be impossible for the destination host to complete
   reassembly of the packet if the first fragment is missing, and
   therefore the entire packet will be discarded.

   The Internet Protocol allows fragmentation of packets into pieces so
   small as to be impractical because of data and computational
   overhead.  Attackers can sometimes exploit typical filter behavior
   and the ability to create peculiar fragment sequences in order to
   sneak otherwise disallowed packets past the filter.  In normal
   practice, such pathalogical fragmentation is never used, so it is
   safe to drop these fragments without danger of preventing normal
   operation.






Ziemba, Reed & Traina        Informational