[misc.security] "breaking" CRCs

jimkirk@OUTLAW.UWYO.EDU (James Kirkpatrick) (02/24/91)

I need/want some specific references, examples, or descriptions of how
to "break" a CRC-based checksumming system.  The specific case is VMS'
undocumented (but widely known and used) CHECKSUM command which reads the
contents of a given file and produces a checksum.  This is then used to
verify that the file "probably" was not altered (e.g. patched by some
unauthorized person).  I'm aware that CRC-based checksumming is not
nearly as secure as, say, DES-based hashing.  I've been told that if
the CRC polynomial is known and I can change the last "n" bits of the
file (where "n" is the size of the CRC), it is easy to fudge the file
so the checksum matches again.  Otherwise it is more difficult.

I'd like some specific references, if at all possible, or examples.
My manager needs convincing that cryptographic checksums should be
used as opposed to CRC-based checksums, and I'd really like to be
able to demonstrate the weaknesses.

Jim Kirkpatrick     JIMKIRK@CORRAL.UWYO.EDU   JIMKIRK@UWYO (Until 6/1)